11.05.2016 Views

TOMBO Ver.2 Manual

TOMBO

TOMBO

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>TOMBO</strong> <strong>Ver.2</strong> <strong>Manual</strong><br />

for LDA and GW Calculations<br />

30 October, 2015<br />

<strong>TOMBO</strong> Group


Preface<br />

This <strong>Manual</strong> is prepared for all users, who want to perform LDA and GW calculations<br />

with <strong>TOMBO</strong> <strong>Ver.2</strong>. Chapters 1-5 describe a basic concept of <strong>TOMBO</strong>, a basic knowledge<br />

for GW, Input files, Output files, and an example of GW calculation of rutile TiO 2 ,<br />

respectively. In Appendix A, usage of the user interface “Materials Studio®” for crystal<br />

calculations is described. <strong>TOMBO</strong> is a first-principles program using the all-electron mixed<br />

basis approach, in which one-particle (Kohn–Sham or quasiparticle) wave functions are expressed<br />

in a linear combination of both atomic orbitals and plane waves. This approach<br />

is described in detail in the following reference. Anyone who publishes any result using<br />

<strong>TOMBO</strong> has to cite this reference:<br />

Shota Ono, Yoshifumi Noguchi, Ryoji Sahara, Yoshiyuki Kawazoe, and Kaoru Ohno,<br />

<strong>TOMBO</strong>: All-electron mixed-basis approach to condensed matter physics ,<br />

Computer Physics Communications 189, 20-30 (2015). DOI: 10.1016/j.cpc.2014.11.012<br />

<strong>TOMBO</strong> can handle both isolated systems and crystal (or periodic) systems in a unified<br />

way. It can perform not only LDA calculations but also GW (+ Bethe–Salpter equation)<br />

calculations.<br />

The <strong>TOMBO</strong> <strong>Ver.2</strong> code, manuals, and related information are available from the web<br />

site of Kaoru Ohno Laboratory at Yokohama National University:<br />

http://www.ohno.ynu.ac.jp/tombo/index.html<br />

Any correspondence can be sent to tombo@ynu.ac.jp.<br />

30 October 2015<br />

<strong>TOMBO</strong> Group


Table of contents<br />

1 Introduction 1<br />

1.1 Short introduction of <strong>TOMBO</strong> . . . . . . . . . . . . . . . . . . . . . . . . 1<br />

1.2 Mixed basis formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 2<br />

1.3 All-electron charge density and potential . . . . . . . . . . . . . . . . . . . 6<br />

1.4 Calculation flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10<br />

1.5 TDDFT dynamics simulation . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />

2 The GW Approximation 13<br />

2.1 The quasiparticle band gap . . . . . . . . . . . . . . . . . . . . . . . . . . 13<br />

2.2 The Green function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br />

2.3 The self-energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15<br />

2.4 Hedin’s equations (GWΓ method) . . . . . . . . . . . . . . . . . . . . . . 16<br />

2.5 The GW approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />

2.6 One-shot GW approximation . . . . . . . . . . . . . . . . . . . . . . . . . 19<br />

3 Input Files 21<br />

3.1 COORDINATES.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />

3.1.1 rct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.1.2 nc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.1.3 nv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.1.4 ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.1.5 np . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.2 INPUT.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

3.2.1 iApp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.2.2 iAlg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.2.3 iexc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.2.4 iCHG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29<br />

3.2.5 Ulevel, Llevel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30


Table of contents<br />

iv<br />

3.2.6 ippmG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

3.2.7 jmax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

3.2.8 deltaE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

3.2.9 npp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br />

3.2.10 nod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br />

3.2.11 nog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31<br />

3.2.12 ncut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />

3.2.13 noz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />

3.2.14 ncs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32<br />

3.2.15 ncc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

3.2.16 nol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

3.2.17 nband . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

3.2.18 q_eliminate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

3.2.19 q_min_mod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

3.2.20 icontinue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />

3.2.21 isphcut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />

3.2.22 iTotalE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />

3.2.23 lpri . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />

3.2.24 nspin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34<br />

3.2.25 irelative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

3.2.26 nion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

3.2.27 iExcite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

3.2.28 nonadiabatic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

3.2.29 iMIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35<br />

3.2.30 smixSCF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.31 nSDCG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.32 iSblp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.33 iSblp2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.34 mSblp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.35 smixTD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

3.2.36 iasym . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

3.2.37 icalAO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

3.2.38 iCry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

3.2.39 precLDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

3.2.40 iFTapp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37<br />

3.2.41 iCheb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37


Table of contents<br />

v<br />

3.2.42 Mcheb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />

3.2.43 nStep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />

3.2.44 singlet, triplet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />

3.3 KPOINT.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38<br />

3.4 QPOINT.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40<br />

3.5 SPOINT.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42<br />

4 Output Files 44<br />

4.1 OUTPUT.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44<br />

4.2 Status.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />

4.3 ISYS.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />

4.4 posit.dat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />

4.5 eigen.dat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />

4.6 MD_Coordinates.xyz, MD_Coordinates.arc . . . . . . . . . . . . . . . . . 45<br />

4.7 band.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45<br />

4.8 WaveF_HOMO-1.cube, WaveF_HOMO-1.grd,<br />

WaveF_HOMO-1.vasp, and so on . . . . . . . . . . . . . . . . . . . . . . 46<br />

4.9 GWA.out . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46<br />

4.10 PhotoAbsorptionSpectra.out . . . . . . . . . . . . . . . . . . . . . . . . . 49<br />

5 Examples 50<br />

5.1 LDA calculation of isolated dimers . . . . . . . . . . . . . . . . . . . . . . 50<br />

5.2 Molecular Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53<br />

5.3 GW calculation of rutile TiO 2 . . . . . . . . . . . . . . . . . . . . . . . . 55<br />

5.4 LDA calculation of rutile TiO 2 . . . . . . . . . . . . . . . . . . . . . . . . 56<br />

5.5 Wave function calculation of rutile TiO 2 . . . . . . . . . . . . . . . . . . . 59<br />

References 61<br />

Appendix A Materials Studio 63<br />

A.1 Build crystal model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64<br />

A.2 1 st Brillouin zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65<br />

A.3 *.cell file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66<br />

A.3.1 make *.cell file . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66<br />

A.3.2 information of *.cell . . . . . . . . . . . . . . . . . . . . . . . . . 68


Chapter 1<br />

Introduction<br />

<strong>TOMBO</strong>[1, 2], all-electron mixed-basis code, is short for "TOhoku Mixed Basis Orbitals<br />

ab initio program". In this thesis, electronic structures of transition metal oxides are<br />

calculated by our original all-electron mixed basis code <strong>TOMBO</strong> developed by Prof. Kaoru<br />

Ohno, et al..<br />

1.1 Short introduction of <strong>TOMBO</strong><br />

Density funcitonal theroy (DFT) [3] and local density approximation (LDA) [4] have<br />

been used in numerous electronic structure calculations and firt-principles molecular dynamics<br />

(MD) simulations. To perform those calculations, Kohn–Sham (KS) equation needs<br />

to be solved self-consistently such that the input potential is identical to the output potential<br />

[4]. The electronic wave function has to be described by appropriate functions to solve<br />

the KS equation. Among the plane wave (PW) expansion approach has been applied to the<br />

ab-initio molecular dynamics (MD) simulations with reasonably high accuracy [5].<br />

However, it is difficult to treat phenomena (hyperfine interaction, XPS, XANES, etc.)<br />

related to core electrons by this method. One problem in generating good pseudo potentials<br />

is related to the fact that the core contribution to the exchangecorrelation potential is not<br />

simply an additive quantity. Moreover, it is not easy to create efficient pseudopotentials,<br />

which require only small number of plane waves.<br />

On the other hand, linear combination of atomic orbitals (LCAO) approaches can treat<br />

all electrons. However, these methods have an intrinsic problem of incomplete basis set, and<br />

therefore there is a problem in applying them in perturbation theory or spectral expansion,<br />

which requires a description in the complete Hilbert space. It is also difficult to consider a<br />

negative affinity problem by these methods. Related but slightly different problem inherent


1.2 Mixed basis formulation 2<br />

to these methods is a basis set superposition error (BSSE) [6, 7]. There is also some trouble<br />

in the Gaussian basis method [8] to describe cusp in the wave-funcion at the nuclear position.<br />

In these respects, it is highly desirable to develop new method, which combines the<br />

PW expansion technique with the LCAO technique to remove pseudopotentials in the PW<br />

expansion methods and to make the basis set complete in the LCAO methods. This is the<br />

main idea to introduce the all-electron mixed basis approach. <strong>TOMBO</strong> [1, 2] is the program<br />

package using this approach. <strong>TOMBO</strong> is the all-electron first-principles method, which can<br />

be applicable to both isolated and periodic systems with complete basis set. It is not the<br />

overcomplete basis set because only limited number of PWs is used in the computation.<br />

The powerfulness of <strong>TOMBO</strong> is not only based on these features but also based on the<br />

fact that it enables us to perform the state-of-the-art calculations such as GW approximation<br />

and Bether–Salpeter equation (BSE). Using these methods, <strong>TOMBO</strong> can treat the problems<br />

related to electron correlations, electronic structure around the band gap, excitation spectra,<br />

and so forth. This <strong>TOMBO</strong> <strong>Ver.2</strong> is a new version made by unifying preexiting several<br />

versions: Molecular dynamics code [9], crystal code [10], T -matrix and GW + BSE code<br />

[11], and <strong>TOMBO</strong> Ver.1 [12].<br />

1.2 Mixed basis formulation<br />

The “mixed-basis” indicates the method using both plane waves and Bloch sums made<br />

of atomic orbitals as the basis set.<br />

For an isolated atom, it is possible to solves the Kohn–Sham equation very rigorously,<br />

because of the system has a spherical symmetry. In this case, the Kohn-Sham wave function<br />

is expressed as a product of radial function R jnl (r) and spherical harmonics Y lm (r) as<br />

ϕ AO<br />

jnlm (r) = R jnl(r)Y lm (ˆr) (1.1)<br />

Here, j, n, l, and m are atomic species, principal quantum number, angular momentum<br />

quantum number, and magnetic quantum number.<br />

In the mixed basis code, the KS wave function is expressed as a linear combination of<br />

PWs and AOs (Fig. 1.1),<br />

ψ v (r) = √ 1<br />

Ω<br />

∑<br />

c PW<br />

G<br />

v (G)e i(k+G)·r +∑<br />

j<br />

∑<br />

nlm<br />

c AO<br />

v<br />

( jnlm)ϕ Bloch (k,r) (1.2)<br />

jnlm<br />

with<br />

ϕ Bloch<br />

jnlm (k,r) = eik·R j<br />

∑<br />

R<br />

e ik·R ϕ AO<br />

jnlm (r − R j − R). (1.3)


1.2 Mixed basis formulation 3<br />

Here, Ω is the volume of the unit cell, G is the reciprocal lattice vector, c is the expansion<br />

coefficients. Any system must be defined inside the unit cell, which is periodic in space (Fig.<br />

1.2). The (valence) radial function R jnl (r) is truncated by subtracting a smooth polynomial<br />

function that satisfies the matching condition at the surface of atomic sphere, while the rest<br />

smooth function (the polynomial function inside and the tail outside atomic sphere) can be<br />

expressed by PWs; see Fig. 1.3.<br />

Fig. 1.1 In the all-electron mixed basis approach, one-electron wave functions are expressed<br />

in a linear combination of plane waves (PWs) and atomic orbitals (AOs).<br />

Fig. 1.2 Any system is defined inside the unit cell, which is periodic in space.<br />

Fig. 1.3 Each AO is confined inside the non-overlapping atomic sphere by subtracting a<br />

smooth polynomial function.<br />

Usually, AOs are generated at the outset by the atomic LDA code base on a modified<br />

Herman–Skillman’s algorithm and not updated during the calculation. However, if one<br />

wants to update AOs during the self-consistent field (SCF) loop, one can set icalAO = 1 in<br />

INPUT.inp. This option is available for cluster calculations only.


1.2 Mixed basis formulation 4<br />

Many-body perturbation theory (such as the GW calculations) or spectral method (such<br />

as the expansion of the wave packet in terms of the eigenstates of the Hamiltonian in the<br />

TDDFT dynamics simulations) requires summing over large number of empty states. The<br />

PW basis set can most accurately describe the empty states. In contrast, to describe the<br />

electrons in the core region accurately, the AO basis set works better then the PW basis<br />

set. The all-electron mixed basis approach, using both PWs and AOs as a basis set in a<br />

combined way, is able to meet the requirements to describe both spatially extended and<br />

localized states. In our code, AOs are numerically described inside the non-overlapping<br />

atomic spheres and the radial part is treated using the logarithmic mesh.<br />

The all-electron mixed basis approach has the following advantages:<br />

1. The number of basis functions can be significantly reduced.<br />

2. In Hamiltonian matrix elements, it is not necessary to store PW-PW part because it is<br />

given simply by the Fourier components V (G − G ′ ).<br />

3. It is possible to accurately treat core states because we determine AOs by using<br />

Herman–Skillman code with logarithmic radial mesh.<br />

4. There is no complexity to generate and treat pseudoptentials. There is also no problem<br />

of transferability.<br />

5. The overlap between AOs and PWs is calculated accurately by first performing angular<br />

integral analytically and then performing radial integral of spherical Bessel functions<br />

numerically in logarithmic radial mesh.<br />

6. Because AOs are confined inside non-overlapping atomic spheres, there is no BSSE<br />

problem, and it is not necessary to calculate overlap integrals between AOs centered<br />

at different atoms, which might produce unnecessary computational errors. Simultaneously,<br />

this reduces the overcompleteness problem.<br />

Since the basis functions (PWs and AOs), f ξ (r), are not orthogonal each other, the<br />

eigenvalue ε ν is obtained by solving the following generalized eigenvalue problem:<br />

∑H ξ ξ ′c ν,ξ ′ = ε ν ∑S ξ ξ ′c ν,ξ ′, (1.4)<br />

ξ ′ ξ ′<br />

where H ξ ξ ′ = ⟨ f ξ |H| f ξ ′⟩ and S ξ ξ ′ = ⟨ f ξ | f ξ ′⟩ are, respectively, the Hamiltonian and overlap<br />

matrix elements between the ξ ’th and ξ ′ ’th basis functions. They looks like<br />

H =<br />

(<br />

⟨PW|H|PW⟩<br />

⟨AO|H|PW⟩<br />

⟨PW|H|AO⟩<br />

⟨AO|H|AO⟩<br />

)<br />

, S =<br />

(<br />

⟨PW|PW⟩<br />

⟨AO|PW⟩<br />

⟨PW|AO⟩<br />

⟨AO|AO⟩<br />

)<br />

. (1.5)


1.2 Mixed basis formulation 5<br />

There are N w PWs (the number of plane waves "nw" in the code), N ao AOs (the number of<br />

atomic orbitals "nao" in the code), and altogether N bs basis functions (the number of basis<br />

set "nbs" in the code). The corresponding column vector is written as<br />

⎛<br />

Ψ ν ≡<br />

⎜<br />

⎝<br />

Eq.(1.4) is then rewritten as follows:<br />

c ν,1<br />

c ν,2<br />

.<br />

c ν,Nw<br />

c ν,Nw +1<br />

.<br />

C ν,Nbs<br />

⎞<br />

⎟<br />

⎠<br />

⎫<br />

1st G<br />

2nd G<br />

⎪⎬<br />

“nw”<br />

.<br />

last G<br />

⎪⎭<br />

(1.6)<br />

⎫<br />

1st AO ⎪⎬<br />

.<br />

⎪⎭ “nao” last AO<br />

HΨ ν = ε ν SΨ ν . (1.7)<br />

This generalized eigenvalue problem is transformed to the usual eigenvalue problem by<br />

using Choleski decomposition. The overlap matrix S is expressed as a product of a lower<br />

triangular matrix T and its Hermitian conjugate T † :<br />

S = T T † . (1.8)<br />

Then, Eq. (1.7) becomes<br />

H ′ Φ ν = ε ν Φ ν , (1.9)<br />

with the transformed Hamiltonian<br />

H ′ = T −1 HT †−1 , (1.10)<br />

(H ′ is calculated from H and T −1 ), and the transformed eigenvectors are give by<br />

Φ ν = T † Ψ ν . (1.11)<br />

The resulting ordinary eigenvalue problem can be solved by the standard library program<br />

using the Householder method, if iApp = D is assigned in INPUT.inp.<br />

In order to calculate the Hamiltonian and Overlap matrix elements, PW-PW, PW-AO,<br />

and AO-AO parts are calculated independently. Overlap matrix between two PWs (G,G ′ )<br />

is obviously a unit matrix. Similarly, Hamiltonian matrix between two PWs is the same as


1.3 All-electron charge density and potential 6<br />

the usual PW expansion method:<br />

⟨k + G|H|k + G ′ ⟩ = ¯h2 (k + G) 2<br />

δ<br />

2m GG ′ +Ṽ (G − G ′ ). (1.12)<br />

PW-AO matrix elements can be accurately calculated by 1D integration of the product of<br />

the spherical Bessel function j l ′(Gr) and the radial AO function R AO<br />

jnl (r):<br />

⟨k + G|H|ϕ Bloch e−iG·R ∫<br />

jnlm ⟩ = j<br />

√ Y lm ( k + ˆ G)<br />

Ω cell<br />

[ ¯h 2 (k + G) 2 ]<br />

j l (|k + G|r)<br />

+V (r)<br />

2m<br />

× R jnl (r)r 2 dr. (1.13)<br />

AO-AO matrix elements can be accurately calculated by 1D integration of the product of<br />

two radial AO functions:<br />

∫ [<br />

⟨ϕ Bloch Bloch<br />

jnlm |H|ϕ j ′ n ′ l ′ m ′⟩ = δ j j ′ ϕ AO<br />

jnlm (r) − ¯h2<br />

cell 2m ∇2 +V (r)<br />

]<br />

ϕ A0<br />

jn ′ l ′ m ′(r)d3 r. (1.14)<br />

The overlap matrix elements are given by those of (1.13) and (1.14) without [...] inside the<br />

integrands.<br />

1.3 All-electron charge density and potential<br />

In the all-electron mixed basis approach, all-electron charge density ρ(r) is made of<br />

three contributions: PW-PW, AO-PW,and AO-AO.<br />

ρ(r) = ρ PW−PW (r) +∑ρ AO−PW<br />

j<br />

j (r) +∑<br />

j<br />

ρ AO−AO<br />

j (r). (1.15)<br />

In the all-electron mixed basis approach, the charge density is made of the three contributions,<br />

ρ AO−PW<br />

j (r) = √ 2 occ<br />

Ω<br />

occ<br />

j (r) = 2<br />

ρ AO−AO<br />

∑<br />

v<br />

ρ PW−PW (r) = 2 Ω<br />

∑ ∑<br />

v<br />

nlm<br />

∑ ∑<br />

n ′ l ′ m ′ nlm<br />

∑<br />

G<br />

occ<br />

∑<br />

v<br />

∑∑<br />

G<br />

c PW∗<br />

v<br />

G ′<br />

Here the prefactor 2 denotes the spin duplicity, ∑ occ<br />

ν<br />

(G ′ )c PW<br />

v (G)e i(G−G′ )·r , (1.16)<br />

c AO∗<br />

v ( jnlm)c PW<br />

v (G) × ϕ jnlm (r − R j )e i(G)·r + c.c., (1.17)<br />

c AO∗<br />

v ( jn ′ l ′ m ′ )c AO<br />

v ( jnlm) × ϕ jn ′ l ′ m ′(r − R j)ϕ jnlm (r − R j ). (1.18)<br />

means the sum over all occupied states


1.3 All-electron charge density and potential 7<br />

(in a case of non-spin-polarized systems), and c.c. in equation (1.17) means the complex<br />

conjugate of the previous term. The first PW-PW contribution can be conveniently treated<br />

in Fourier space. The rest two AO-related contributions are confined only inside the nonoverlapping<br />

atomic spheres, and can be written together as<br />

ρ AO<br />

j<br />

(r) = ρ AO−PW<br />

j<br />

(r)ρ AO−AO<br />

j (r). (1.19)<br />

This can be divided into two parts: one is spherical symmetric part σ j (r j ) and the other is<br />

asymmetric part. Here we put r j = |r − R j |. The asymmetric part is generally negligible,<br />

but if necessary can be taken into account by setting the option parameter iAsym = 1 in<br />

INPUT.inp. The treatment of ignoring asymmetric part of the AO-related charge density<br />

is guaranteed when the radii of non-overlapping spheres are set reasonably small enclosing<br />

the core region only.<br />

Spherical potential V j (r j ) inside each non-overlapping atomic sphere is calculated as<br />

follows. First, consider the Hartree potential made by the AO-related spherically symmetric<br />

charge σ j (r j ) centered at R j . This potential is easily calculated as the 1D integration in<br />

radial direction of the Poisson equation as follows:<br />

v H j (r j ) = 4π<br />

r j<br />

∫ r j<br />

0<br />

∫ rc<br />

σ j (r)r 2 dr + 4π σ j (r)rdr. (1.20)<br />

r j<br />

According to the Gauss theorem in electrostatics, this Hartree potential behaves as<br />

v H j (r j ) = Q j<br />

r j<br />

, for r ≥ r c , (1.21)<br />

where Q j is the symmetric charge defined as<br />

∫ rc<br />

Q j = 4π σ j (r)r 2 dr. (1.22)<br />

0<br />

On the other hand, if we define the screened charge Z ∗ j as<br />

Z ∗ j = Z j − Q j , (1.23)<br />

(Z j is atomic number), the sum of the Hartree potential v H j (r j) and nuclear Coulomb potential<br />

−Z j /r j around R j becomes<br />

v H j (r j ) − Z j<br />

r j<br />

= − Z∗ j<br />

r j<br />

, for r j ≥ r c (1.24)<br />

outside the non-overlapping atomic sphere. Because of this long tail, it is necessary to add


1.3 All-electron charge density and potential 8<br />

all the contributions from surrounding atoms. This summation should be taken not only<br />

inside the own unit cell, but also surrounding or further apart unit cells. To treat this accurately,<br />

we use the following Fourier decoupling method. In this method, the simple potential<br />

form −Z ∗ j /r j given by Eq.(1.24) connects smoothly to the quadratic function inside the nonoverlapping<br />

atomic sphere. They should have the same value and the same derivative at the<br />

radius of the atomic sphere r = r c :<br />

⎧<br />

⎨<br />

v interpo Z ∗ j<br />

j<br />

(r j ) =<br />

(br2 j + d) for r j < r c ,<br />

⎩−Z ∗ j /r j for : r j ≥ r c .<br />

From the matching condition, we obtain<br />

(1.25)<br />

br 2 c + d = −1/r c , 2br c = 1/r 2 c. (1.26)<br />

From simple calculation, we find that these conditions are identical to<br />

b = 1<br />

2rc<br />

3 , d = − 3 . (1.27)<br />

2r c<br />

Thus connected potential is a smooth and analytic function over whole space and is easily<br />

transformed into reciprocal lattice space analytically. We call this potential the interpolated<br />

Coulomb potential and write it as v interpo<br />

j<br />

(r j ); see Eq.(1.25). This interpolated Coulomb<br />

potential takes the correct value of the Coulomb potential, v H j (r j) − Z j<br />

r j<br />

, everywhere outside<br />

the jth atomic sphere; It takes an incorrect value only inside the jth atomic sphere, and its<br />

difference from the correct value is given by<br />

Vj<br />

trunc (r j ) = v H j (r j ) − Z j<br />

− v interpo<br />

j<br />

(r j ), (1.28)<br />

r j<br />

which we call the truncated Coulomb potential. It has nonzero values only each atomic<br />

sphere, and is spherically symmetric. This truncation is schematically illustrated in Fig.1.4.<br />

This truncated Coulomb potential is added to the truncated spherical exchange-correlation<br />

potential µ xc<br />

j (|r − R j |) and stored as one-dimensional data on the radial logarithmic mesh.<br />

Apart from this truncated function, we have to treat separately interpolated Coulomb<br />

potential v interpo<br />

j<br />

(r j ). This function v interpo<br />

j<br />

(r j ) is analytically expressed by Eq.(1.25) in infinite<br />

space, and therefore is able to be analytically transformed into G space. The Fourier<br />

coefficients are explicitly given by<br />

{<br />

¯v interpo<br />

j<br />

(G) = 4πZ je −iG·R j<br />

b [( 3(Gr c ) 2 − 6 ) ]<br />

sinGr c + 6Gr c cosGr c /G<br />

5<br />

Ω


1.3 All-electron charge density and potential 9<br />

+d sinGr c /G 3 + cosGr c /G 2 }. (1.29)<br />

This analytic Fourier coefficients are added to the rest part of the Fourier coefficients of<br />

both the Hartree potential ¯ρ rest (G) and the exchange-correlation potential ¯µ xcrest (G) to give<br />

an additional contribution to the spherical potential in the jth atomic sphere. That is, this<br />

additional contribution to the spherical potential from the three kinds of Fourier coefficients<br />

are expressed as:<br />

Vj<br />

rest (r j ) = ∑<br />

G<br />

sinGr j<br />

e iG·R j<br />

Gr j<br />

[<br />

∑<br />

k<br />

¯V interpo<br />

k<br />

(G) + 4π Ω<br />

¯ρ rest (G)<br />

G 2<br />

+ ¯µ xcrest (G)<br />

]<br />

. (1.30)<br />

Here we note that v interpo<br />

j<br />

(r j ) defined by Eq.(1.25) is different from the first term in the l.h.d.<br />

of Eq.(1.30). The latter includes all the tails of the interpolated Coulomb potential centered<br />

at all k ≠ j atoms. This is obvious from the nature of the Fourier transformation.<br />

Thus the total effective potential used in the AO-related matrix elements is given by<br />

V j (r j ) = Vj<br />

trunc (r j ) +Vj rest (r j ). (1.31)<br />

All the AO-related matrix elements are calculated by using this spherical potential. If iCheb<br />

= 1 is set in INPUT.inp, a Chebyshev polynomial fitting to this spherical potential inside<br />

atomic sphere is used to accelerate the computational speed. The contribution from the<br />

asymmetric part of the potential is treated separately in a large Fourier space when iAsym =<br />

1 is set in INPUT.inp.<br />

Fig. 1.4 How to calculate periodic Hartree potential V H (r) inside non-overlapping atomic<br />

spheres?


1.4 Calculation flow 10<br />

1.4 Calculation flow<br />

Flowchart of the first-principles molecular dynamics (MD) is shown in Fig. 1.5.<br />

Set initial atomic positions<br />

❄<br />

Assume initial charge density<br />

❄<br />

Initialize coefficients C λ i<br />

✲<br />

❄<br />

Calculate charge density<br />

❄<br />

Mix the charge density with<br />

that of the previous step<br />

Calculate the KS potential<br />

❄<br />

Calculate matrix ❄ elements<br />

❄<br />

Update Ci<br />

λ according to<br />

matrix diagonalization or<br />

steepest-descent method<br />

❄<br />

NO ✘✘✘✘✘✘✘<br />

✛ ✘❳❳❳❳❳❳❳❳❳❳❳❳❳❳❳❳❳❳<br />

✘ If |E new − E old | < precLDA<br />

✘✘✘✘✘✘✘✘✘<br />

YES<br />

❄<br />

Calculate forces on atoms<br />

❄<br />

Update atomic positions<br />

Check self-consistency<br />

Fig. 1.5 Flowchart of the first-principles MD. The convergence of the electronic states is<br />

checked by the difference between the total energies at the present and previous steps. If<br />

convergence is achieved, the atomic positions are updated by using the calculated force.


1.5 TDDFT dynamics simulation 11<br />

1.5 TDDFT dynamics simulation<br />

In <strong>TOMBO</strong>, TDDFT dynamics simulation can be performed by setting “A” in iApp in IN-<br />

PUT.inp. If one wants to excite one electron from HOMO to LUMO +n at the outset, one<br />

may define iExcite = 1 + n in INPUT.inp (n can be 0, 1, 2, ...). Default value of iExcite is<br />

0, which means no excitation.<br />

According to the time-dependent density functional theory[13], <strong>TOMBO</strong> can treat the<br />

time-dependent Kohn–Sham (TDKS) equation<br />

i ∂ ∂t ψ j(r,t) = H q (r,t)ψ j (r,t), (1.32)<br />

by setting “A” in iApp in INPUT.inp Here H q (r,t) is the electronic part of the Hamiltonian.<br />

Combining this time-evolution equation with the Newtonian equation of motion for nuclei<br />

M A ¨R A = − ∂<br />

∂R A<br />

[<br />

E q +V cl] , (1.33)<br />

we can perform semi-classical dynamics simulation within the mean-field-type Ehrenfest<br />

theorem. During one simulation, all dynamical variables change in time along one reaction<br />

path according to the choice of the initial atomic coordinates and initial atomic velocities<br />

(velocities can be written parallel to the right of the coordinates in COODINATES.inp). This<br />

way of simulation is called “on the fly” approach. Here E q is the electronic part of the total<br />

energy, and, together with the Coulomb potential between nuclei V cl , gives the mean-field<br />

potential exerting on nuclei. In these equations, r is the electron coordinate, M A and R A are<br />

the mass and position of the Ath nucleus. This approach is on the fly.<br />

The simulation is basically performed adiabatically, i.e., the atomic motion is assumed<br />

slow enough. However, non-adiabatic simulation is possible by considering the coupling to<br />

the atomic velocity, when the input parameter nonadiabatic = 1 is set in INPUT.inp. Below<br />

only the algorithm of adiabatic simulation is explained, but one may refer to Ref. [14] for<br />

non-adiabatic simulation.<br />

It is necessary to know the exchange-correlation potential to solve Eq.(1.32) step by step<br />

by means of the TDDFT. <strong>TOMBO</strong> uses a simple LDA exchange-correlation functional that<br />

is local for both space and time. This approximation is called “adiabatic LDA”. In order to<br />

integrate Eq.(1.32) step by step accurately, we use the spectral method[15], in which wave<br />

packet ψ j (r,t) at each instantaneous time is expanded in terms of eigenfunctions ϕ k (r,t)<br />

(eigenvalues are ε k (t))<br />

H q (r,t)ϕ k (r,t) = ε k (t)ϕ k (r,t). (1.34)


1.5 TDDFT dynamics simulation 12<br />

of the Hamiltonian at that time H q (r,t). Wave packes are expanded as<br />

ψ j (r,t) = ∑c jk (t)ϕ k (r,t), (1.35)<br />

k<br />

where the coefficients c jk (t) are given by the inner product of ϕ k (r,t) and ψ j (r,t) as<br />

c jk (t) = ⟨ϕ k (r,t)|ψ j (r,t)⟩. (1.36)<br />

Therefore, if we set the time step ∆t smaller than the time scale where the Hamiltonian<br />

changes, the TDKS equation can be integrated as follows:<br />

ψ j (r,t + ∆t) = ∑exp[−iε k (t)∆t]c jk (t)ϕ k (r,t). (1.37)<br />

k<br />

When the Hamiltonian does not change in time, Eq.(1.37) is exact. Of course, electron wave<br />

packets oscillate 100-1000 times faster than the nuclear motion, but the stability of Eq.(1.37)<br />

is excellent. Typically if one set ∆t at 10 −2 -10 −1 fs, the Hamiltonian H q (r,t) almost does<br />

not change, Eq.(1.37) becomes good approximation.<br />

This spectral method requires that the basis functions span complete space at least approximately,<br />

and in this sense, the all-electron mixed basis method is suitable. In fact, the<br />

summation over the eigenstates in Eq.(1.37) can be restricted to low lying excited states only,<br />

although continuum free-electron-like states above energy zero should be also included. The<br />

number of levels taken into account in this summation is set as nol in INPUT.inp. Usually,<br />

nol = 500 is recommended, but it should be set 1000 or more for larger systems.<br />

Newtonian equation of motion (1.33) involves forces calculated by the Coulomb force<br />

between nuclei and the derivative of the electron total energy with respect to the nuclear<br />

positions. These forces are calculated in the same way described in previous section. If<br />

the electronic states is on an adiabatic surface, the dynamics using this mean-field potential<br />

becomes the adiabatic molecular dynamics. If the wave packet ψ j (r,t) is the superposition<br />

of the eigenstates, the forces acting on nuclei becomes the average of the forces calculated<br />

from the different eigenstates with the weight of these eigenstates in the wave packet (mixed<br />

state). This is the problem of this mean-field approximation. This mean-filed force becomes<br />

unphysical apart from the region where the mean-field potential is valid.<br />

To do the TDDFT dynamics simulation, it is necessary to first iterate the SCF loop<br />

until self-consistency is obtained. Then the TD dynamics loop starts. Typically dTime<br />

= ∆t = 0.01 ∼ 0.1 [fs] should be put in COORDINATES.inp. As well as the usual MD,<br />

nStep should be also given in INPUT.inp.


Chapter 2<br />

The GW Approximation<br />

The calculated DFT band structures are not improved by the energy-independent nonlocaldensity<br />

corrections to the LDA, in any case, quasiparticle energies are needed [16]. In order<br />

to improve the accuracy of electronic structure calculation of materials, the GW approximation<br />

is introduced on the basis of many-body perturbation theory (MBPT)[17].<br />

2.1 The quasiparticle band gap<br />

In the N-particle many-body neutral system, eigenstates and total energies are denoted<br />

|ψ⟩ and E N , respectively. The electron quasiparticle (QP) energies are defined by<br />

ε ck = E N+1 − E0 N (unoccupied) (2.1)<br />

and the hole QP energies are defined by<br />

ε vk = E0 N − EN−1 (occupied) (2.2)<br />

These are excitation energies of the (N ± 1)-particle system relative to the N-particle<br />

ground-state energy E0 N and thus correspond to the electron addition and removal energies.<br />

It is clear that ε ck > ε F and ε vk ⩽ ε F , where ε F is the Fermi level. The quasiparticle energy<br />

gap is given by<br />

E gap = ε {CBM} − ε [V BM] = E N+1 + E N−1 − 2E0 N (2.3)<br />

where {...} and [...] indicate the energetically minimum and maximum k points of ..., and,<br />

at each k point, the CBM and VBM refer to the conduction band minimum and the valence<br />

band maximum, respectively.


2.2 The Green function 14<br />

2.2 The Green function<br />

The propagation of one particle through the system is described by the one particle<br />

Green function. The one particle Green function is defined as:<br />

∣ [<br />

]∣ ⟩<br />

∣∣T<br />

G(x 1 ,t 1 ,x 2 ,t 2 ) = i<br />

⟨Ψ N ψ(x 1 ,t 1 )ψ † ∣∣ΨN<br />

(x 2 ,t 2 )<br />

(2.4)<br />

It contains the information on energy and lifetimes of the quasiparticle, and also on<br />

the ground state energy of the system and the momentum distribution. Here, x = (x,t) =<br />

(r,σ,t). Ψ N is the Heisenberg ground state vector of the interacting N-electron system satisfying<br />

the engenvalue equation H|Ψ N ⟩ = E|Ψ N ⟩,ψ H and ψ † H<br />

are respectively the annihilation<br />

and creation field operator and T is the Wick time ordering operator.<br />

ψ(x,t) = e iHt ψ(x)e −iHt (2.5)<br />

and the field operator satisfy the anti-commutation relation:<br />

ψ † (x,t) = e −iHt ψ † (x)e iHt (2.6)<br />

{<br />

}<br />

ψ(x),ψ † (x ′ ) = δ(x − x ′ ) (2.7)<br />

and:<br />

{<br />

ψ(x),ψ(x ′ ) } {<br />

}<br />

= ψ † (x),ψ † (x ′ ) = 0 (2.8)<br />

T [ψ(x 1 ,t 1 )ψ † (x 2 ,t 2 )] =<br />

{<br />

ψ(x 1 ,t 1 )ψ † (x 2 ,t 2 ) if t 1 > t 2<br />

ψ(x 2 ,t 2 )ψ † (x 1 ,t 1 ) if t 1 < t 2<br />

(2.9)<br />

Therefore, the Green function describes the probability amplitude for the propagation<br />

of an electron (hole) from position r 2 at t 2 to r 1 at time t 1 for t 1 > t 2 (t 1 < t 2 ) (See Fig.2.1).<br />

Insert a complete set of N+1 and N-1 particle states, ∑ j |ψ N±1 | = 1, we perform a<br />

Fourier transform in energy space we obtain:<br />

j<br />

⟩⟨ψ N±1<br />

j<br />

G(x 1 ,x 2 ;ω) = ∑<br />

s<br />

f s (x 1 fs ∗ (x 2 )<br />

ω − ε s + iηsgn(ε s − ε F )<br />

(2.10)


2.3 The self-energy 15<br />

(a) add an electron<br />

(b) remove an electron<br />

Fig. 2.1 Add/Remove an electron in the N-particle many-body system<br />

where ε F is the Fermi level, and the QP energies are given by<br />

ε s =<br />

{<br />

E s<br />

(N+1) − E (N)<br />

N<br />

− E (N−1)<br />

E (N−1)<br />

N<br />

for ε s ≥ ε F<br />

(2.11)<br />

s for ε s < ε F<br />

The subscript s indicate the quantum label of the states of the N ±1 system. The amplitudes,<br />

i.e., the QP wave functions, f s (x) are defined as:<br />

f s (x) =<br />

{<br />

⟨Ψ N |ψ(x)|Ψ N+1,s ⟩ for ε s ≥ ε F<br />

⟨Ψ N−1,s |ψ(x)|Ψ N ⟩ for ε s < ε F<br />

(2.12)<br />

The Green function has the poles at the electron addition (removal) energies and describes<br />

quasiparticles. So we can obtain the quasiparticle band gap information from Green<br />

function.<br />

2.3 The self-energy<br />

From the definition of Green function (G) and the equation of motion for ψ(r,t) :<br />

i ∂ ∂t<br />

ψ(r,t) = [ψ(r,t),H]. we can show:<br />

i ∂<br />

∫<br />

G(1,2) = δ(1,2) + H 0 (1)G(1,2) − i<br />

∂t 1<br />

ν(1 + ,3)G(1,3,2,3 + )d3 (2.13)


2.4 Hedin’s equations (GWΓ method) 16<br />

Here, H 0 = T +V ext +V H , the number n stands for (r n ,t n ), ν(1 + ,3) is the Coulomb interaction<br />

between electrons. There exists 2-particle Green function, describes the motion of 2<br />

particles:<br />

⟨<br />

[<br />

]∣ ⟩<br />

G(1,2,3,4) = − Ψ N ∣<br />

0 ∣T ψ(1)ψ(2)ψ † (4)ψ † ∣∣Ψ N<br />

(3) 0<br />

In frequency Fourier space:<br />

(2.14)<br />

∫<br />

[ω − H 0 ]G(ω) + i<br />

νG 2 (ω) = 1 (2.15)<br />

Instead of introducing a 2-particles Green function, we introduce the self-energy operator,<br />

Σ . The self-energy allows to close formally the hierarchy of equations of motion of<br />

higher order Green functions. Equation (2.15) is tranformed to 1-particle Green function.<br />

∫<br />

[ω − H 0 ]G(ω) + i<br />

Σ(ω)G(ω) = 1 (2.16)<br />

The self-energy is a non-local and energy dependent operator. From the equation of motion:<br />

∫<br />

[¯hω − H 0 (r)]G(r,r ′ ;ω) −<br />

Σ(r,r ′′ ;ω)G(r ′′ ,r ′ ;ω)d 3 r ′′ = δ(r − r ′ ) (2.17)<br />

Introducing the Lehmann representation for G. The QP energies and QP wave functions<br />

can be obtained as solutions of a Schrödinger-type QP equation [18]<br />

∫<br />

H o ψ nk (r) +<br />

dr ′ Σ(r,r ′ ;ε nk )ψ nk (r ′ ) = ε nk ψ nk (r) (2.18)<br />

2.4 Hedin’s equations (GWΓ method)<br />

It is possible to calculate energy and lifetimes of quasiparticle excitation solving Eq.<br />

(2.18). A formally exact way of calculating the self-energy is given by a set of coupled<br />

equations,known as Hedin’s equations [19]:<br />

Self-energy:<br />

Screened potential:<br />

∫<br />

Σ(1,2) = i<br />

d(34)G(1,3)Γ(3,2,4)W(4,1 + ) (2.19)<br />

∫<br />

W(1,2) = ν(1,2) +<br />

d(34)ν(1,3)P(3,4)W(4,2) (2.20)<br />

Polarization:<br />

∫<br />

P(1,2) = −i<br />

d(34)G(1,3)G(4,1 + )Γ(3,4,2) (2.21)


2.5 The GW approximation 17<br />

Vertex function:<br />

∫<br />

Γ(1,2,3) = δ(1,2)δ(1,3) +<br />

Dielectric function:<br />

d(4567) ∂Σ(1,2) G(4,6)G(7,5)Γ(6,7,3) (2.22)<br />

∂G(4,5)<br />

ε = 1 − νP (2.23)<br />

The screened potential W can also be written as<br />

∫<br />

W(1,2) =<br />

d(3)ε −1 (1,3)ν(3,2) (2.24)<br />

From Hedin’s equations, we know how to calculate the self-energy. Although the GWΓ<br />

method is implemented in a future version of <strong>TOMBO</strong>, the following GW approximation is<br />

needed for the moment.<br />

2.5 The GW approximation<br />

A practical approximation to calculate Σ is the GW approximation proposed by Hedin<br />

[17]. In this GW approximation, the vertex function Γ is approximated in its zeroth-order<br />

form, Γ(1,2,3) = δ(12)δ(13). So the Hedin’s equations become:<br />

Σ(1,2) = iG(1,2)W(1 + ,2) (2.25)<br />

∫∫<br />

G(1,2) = G 0 (1,2) +<br />

d(3)d(4)G 0 (1,3)Σ(3,4)G(4,2) (2.26)<br />

P(1,2) = −iG(1,2)G(2,1) (2.27)<br />

∫∫<br />

W(1,2) = ν(1,2) +<br />

d(3)d(4)ν(1,3)P(3,4)W(4,2) (2.28)<br />

This self-consistent GW calculation is possible in <strong>TOMBO</strong> by setting the first character<br />

of iApp as “G” in INPUT.inp, although crystal calculation and numerical ω integration<br />

(which will be explained below) are not available now.<br />

The explicit form of the self-energy Σ in the GW approximation is given by<br />

Σ GW (r,r ′ ;ε) = i ∫<br />

2π<br />

dωe −i0+ω G(r,r ′ ;ε − ω)W(r,r ′ ;ω) (2.29)


2.5 The GW approximation 18<br />

The GW self-energy can be separated into two terms ( Σ GW = Σx + Σ c ). The exchange part<br />

is given by<br />

Σ x (r,r ′ ) = i ∫<br />

2π<br />

dωe iω0+ G(r,r ′ ;ω)ν(r − r ′ ) = −<br />

occ<br />

∑<br />

nk<br />

ψ nk (r)ψ ∗ nk (r)<br />

|r − r ′ |<br />

(2.30)<br />

In equation (2.30), the symbol occ in the sum means that the summation is taken over the<br />

occupied states only. The diagonal matrix elements of this exchange part of the self-energy<br />

become<br />

⟨<br />

n,k<br />

∣ ∣Σx (r,r ′ ) ∣ ∣ n,k<br />

⟩<br />

= −<br />

occ<br />

∑<br />

∑<br />

n ′ q<br />

∑<br />

G<br />

4π<br />

⟨<br />

ΩG 2<br />

∣<br />

n,k∣e i(q+G)·r∣ ⟩⟨<br />

∣ ∣n ′ ,k − q n ′ ∣<br />

k − q∣e −i(q+G)·r′∣ ⟩<br />

∣ ∣n,k<br />

(2.31)<br />

The correlation part, Σ c can be evaluated by using generalized plasmon-pole (GPP)<br />

model [19]. Σ c can also be evaluated by using the full ω integration [20]:<br />

Σ c (r,r ′ ;ω) = i ∫<br />

2π<br />

dω ′ e −iω0+ G(r,r ′ ;ω − ω ′ ) [ W(r,r ′ ;ω ′ ) − ν(r − r ′ ) ] (2.32)<br />

In Equation (2.32), it is difficult to perform the ω ′ integral along the real axis, since W and<br />

G have a strong structure on this axis. In order to avoid this difficulty, Godby et al.[21–24]<br />

restricted the values of ω to small imaginary numbers and changed the contour of the ω ′<br />

integral from real axis to imaginary axis [16]. Then, by analytic continuation, the resulting<br />

Taylor series is used to estimate the matrix elements for real values of ω. Ishii and Ohno et<br />

al. [20, 25] suggested that this intergration method employed by Godby et al. [16] can be<br />

extended rather easily to real number of ω by slightly modifying the contour. The contour<br />

along the real ω ′ axis from −∞ to +∞ for the integral Equation (2.32) can be replaced by<br />

the contour C shown in Fig.2.2.<br />

Here we further use the symmetry W(ω) = W(−ω) to reduce the contour to the positive<br />

real and imaginary parts only. The diagonal matrix elements of the correlation part of the<br />

self-energy become<br />

⟨<br />

n,k<br />

∣ ∣Σc (r,r ′ ;ω) ∣ ∣ n,k<br />

⟩<br />

= ∑<br />

∑<br />

n ′ q<br />

× i<br />

2π<br />

⟨<br />

∑<br />

G,G ′<br />

∫<br />

∣<br />

n,k∣e i(q+G)·r∣ ⟩⟨<br />

∣ ∣n ′ ,k − q n ′ ∣<br />

k − q∣e −i(q+G′ )·r ′∣ ⟩<br />

∣ ∣n,k<br />

dω ′ [ W G,G ′(q,ω ′ ) − (4π/ΩG 2 ]<br />

)δ G,G ′<br />

1<br />

× (<br />

ω + ω ′ +<br />

− ε k−q,n − iδ k−q,n<br />

1<br />

ω − ω ′ − ε k−q,n − iδ k−q,n<br />

)<br />

(2.33)


2.6 One-shot GW approximation 19<br />

Fig. 2.2 Contour C of the ω ′ integration in Equation 2.33<br />

in the case when we only perform the contour integration along the positive (real and imaginary)<br />

parts of the contour C shown in Fig.2.2 with the help of W(ω) =W(−ω). If we use the<br />

original contour C without using W(ω) = W(−ω), the second term in the parentheses of the<br />

integrand in Equation (12) does not appear. This contour is justified because there is no pole<br />

inside two quadrants of the complex ω ′ plane, corresponding to (Reω ′ > max(ε VBM −ω,0),<br />

Imω ′ > 0) and (Reω ′ < min(ε CBM − ω,0), Imω ′ < 0). Here we used the fact that W(ω ′ )<br />

has no pole in (Reω ′ > 0, Imω ′ > 0) and (Reω ′ < 0, Imω ′ < 0). This term represents the<br />

contribution related to the electron correlation.<br />

Instead of solving equation (2.18) self-consistently, we adopt here the so-called one shot<br />

GW approximation first proposed by Hybertsen and Louie [19]. The Green’s function (G)<br />

is replaced by the LDA Green’s function G 0 and constructed in a non-self-consistent way<br />

from the KS wave functions ψnk KS (r) and eigenvalues εKS<br />

nk<br />

(r). It will be discussed in the next<br />

section.<br />

2.6 One-shot GW approximation<br />

In practice, Kohn-Sham orbitals and eigenvalues from a DFT calculation are often used as<br />

input for a GW calculation and the quasiparticle spectrum is evaluated non-selfconsistenly<br />

from Eq.(2.33) without updating the Green’s Function or the screened potential, that means<br />

only one iteration is made. This is known as the “one-shot” GW or G 0 W 0 approximation<br />

[16, 19] and has become a standard tool in electronic structure theory. W 0 is hereby equal to<br />

the RPA screened potential.<br />

Using the one-shot GW approximation, the quasiparticle equation (Eq. 2.18) is solved<br />

approximately. The quasiparticle spectrum is calculated with Eq. (2.18) using the first-


2.6 One-shot GW approximation 20<br />

order perturbation theory in ( Σ GW −Vxc<br />

LDA ), where Vxc<br />

LDA is the LDA exchange-correlation<br />

potential. Usually, the KS wave functions ψnk KS are sufficiently close to the true quasiparticle<br />

wave function ψ nk (r), so that the first-order estimate of the self-energy correction to the<br />

LDA eigenvalues in adequate [19]. The GW quasiparticle energy εnk<br />

GW is then obtained as<br />

⟨<br />

⟩<br />

= εLDA nk + Z nk ψnk<br />

LDA ∣<br />

∣Σ GW (r,r ′ ;εnk<br />

LDA LDA<br />

) −Vxc (r)δ(r − r ′ ∣<br />

) ∣ψnk<br />

LDA (2.34)<br />

ε GW<br />

nk<br />

with a renormalization factor as<br />

Z nk =<br />

[<br />

1 − ∂Σ ]<br />

GW (ω) −1<br />

∂ω ω=εnk<br />

LDA<br />

(2.35)<br />

The Scheme of the one-shot GW is shown in Fig.(2.3). Firstly, Kohn-Shan equation<br />

is solved by DFT and LDA, it obtain Kohn-Sham eigenvalues and wave fuctions. Then<br />

using one-shot GW approximation, we can get polarization, dielectric function, screened<br />

potential, et. al.. From Hedin’s equations, we obtain the self-energy. Finally, we calculate<br />

the QP energy by Eq.(2.33).<br />

Fig. 2.3 Scheme of the one-shot GW calculation


Chapter 3<br />

Input Files<br />

To perform a LDA or GW calculation of an isolated system (atoms, molecules, or clusters),<br />

<strong>TOMBO</strong> needs two input files, COORDINATES.inp and INPUT.inp. These are the<br />

central input files of <strong>TOMBO</strong>. The variables related to the unit cell, atomic positions, and<br />

also AOs are written in COORDINATES.inp: the lattice constant, lattice vectors, coordinates<br />

of the nuclei and the number of orbital-types of AOs for each atom. On the other<br />

hand, INPUT.inp determines "what to do and how to do it", and can contain a relatively<br />

large number of parameters. In a crystal calculation within a LDA level, two other input<br />

files, SPOINT.inp and KPOINT.inp, are required to assign special k points to do specialpoint<br />

sampling within the irreducible wedge of the 1st Brillouin zone (BZ) during the selfconsistent<br />

field (SCF) loop and output k points to draw band structure diagram, respectively.<br />

Usually, the special points are given on the Monkhorst–Pack grid in the irreducible BZ. For<br />

a GW crystal calculation, the assignment of KPOINT.inp is different from a LDA calculation,<br />

and another input file QPOINT.inp is also required In order to calculate the polarization<br />

function P GG ′ for each momentum transfer q (and for each energy transfer ω in the case of<br />

the full ω integration), k-point sampling is performed for the points on the Gamma grid<br />

in the whole BZ assigned as “sum” in KPOINT.inp. Then the correlation part of the selfenergy,<br />

Σ c is calculated by taking q-point summation on the Γ-grid in the irreducible BZ by<br />

using QPOINT.inp. Finally, the quasiparticle energies as well as the expectation values of<br />

the LDA exchange-correlation potential, and the exchange (Σ x ) and correlation (Σ c ) parts of<br />

the self-energy are evaluated for the k points (typically at symmetry points in the irreducible<br />

BZ) assigned as “out” in KPOINT.inp.


3.1 COORDINATES.inp 22<br />

3.1 COORDINATES.inp<br />

The variables related to both the unit cell and Atomic Orbits (AO) should be written in<br />

COORDINATES.inp: the lattice constant, lattice vectors, coordinates of the nuclei and the<br />

number of orbit-type AOs for each atom. It can also contain the definitions of “mesh” of the<br />

unit cell division (see below) and the basic time step (“dTime” in [fs]) for MD.<br />

A typical form of COORDINATES.inp is shown in Table 3.1.<br />

Table 3.1 Form of COORDINATES.inp<br />

line 1<br />

line 2<br />

name of the target system (arbitrary)<br />

length of lattice vector a 1 in [Å]<br />

line 3 a 1x a 1y a 1z<br />

line 4 a 2x a 2y a 2z<br />

line 5 a 3x a 3y a 3z<br />

line 6 number of atoms<br />

line 7<br />

line 8<br />

line 9<br />

Direct (or Cartesian)<br />

1st atomic species and coordinates<br />

2nd atomic species and coordinates<br />

line 10 ... ...<br />

line 11<br />

line 12<br />

line 13<br />

line 14<br />

line 15<br />

line 16<br />

line 17<br />

nth atomic species and coordinates<br />

XX_rct (radius of atomic sphere of XX atom)<br />

XX_ nc (number of kinds of core states)<br />

XX_nv (number of kinds of valence states)<br />

XX_ns (number of s-type Atomic Orbits for XX atom)<br />

XX_np (number of p-type Atomic Orbits for XX atom)<br />

XX_nd (number of d-type Atomic Orbits for XX atom))<br />

line 18 ... ...<br />

line 19<br />

line 20<br />

mesh (mesh division of unit cell (same for all directions))<br />

END_PARA<br />

Note:<br />

1. Line 1 is arbitrary and can be used for user’s memo of this COORDINATES.inp file.<br />

2. Line 2 is the lattice constant, i.e. the length of the lattice vector a 1 in [Å].


3.1 COORDINATES.inp 23<br />

3. a 1x , a 1y , a 1z are given in line 3, where a 1 should be normalized to 1, i.e., |a 1 | = 1 or<br />

a 2 1x + a 2 1y + a 2 1z = 1. (3.1)<br />

4. a 2x , a 2y , a 2z and a 3x , a 3y , a 3z are given in line 4 and line 5, respectively, in the same<br />

unit of a 1x , a 1y , a 1z .<br />

5. Line 7:"Direct" gives fractional coordinates of atomic positions (coefficients for a 1 ,<br />

a 2 , a 3 ) "Cartesian" gives Cartesian coordinates of atomic positions in [Å] unit.<br />

6. From line 8 to line 11, if "Direct" is chosen, atomic positions should be put around<br />

the center of the unit cell for symmetry. (Note that any coefficients must be between<br />

0 and 1, and cannot be either 0 or 1. Instead of putting 0, you have to put 0.0000001.)<br />

"Cartesian" does it automatically.<br />

7. In order to fix the position of certain atom during the dynamics, F or f should be<br />

written at the right of the atomic coordinate as<br />

Si 0.500000000 0.500000000 0.500000000 F<br />

User can continue MD as a subsequent job by copying the two coordinates at the<br />

last time step and at the time step one before the last step in MD_Coordinates.xyz to<br />

COORDINATES.inp as<br />

C 7.8734002 8.2943019 9.5324198 7.8728956 8.2894257 9.5373841<br />

line by line for all atoms. The coordinates are given in Cartesian in [Å] unit. In this<br />

case, ivelocity = 0 [default] should be set somewhere below line 12.<br />

8. User can set initial velocity of atoms in a similar way:<br />

C 7.8734002 8.2943019 9.5324198 0.010000 0.000000 0.000000<br />

In this case, ivelocity should be set at either 1, 2, or 3 somewhere below line 12. Here,<br />

ivelocity = 1, ivelocity = 2, and ivelocity = 3 mean that the velocity is given in units<br />

of [angstrome/fs], [m/s], and [a.u.], respectively.<br />

9. In order to treat an antiferromagnetic system having zero total magnetic moment, one<br />

should write nspin = -1 in INOUT.inp and simultaneously, one may write the spin<br />

magnetic moment of each atom as m 1 or m -1 at the right of the atomic coordinate as<br />

Mn 0.250000000 0.250000000 0.250000000 m 1<br />

Mn 0.750000000 0.750000000 0.750000000 m -1


3.1 COORDINATES.inp 24<br />

10. The Atomic Orbitals implemented in <strong>TOMBO</strong> code are listed in Fig. 3.1: s-, p-, and<br />

d-orbitals are implemented, while f -orbital is still under implementation. Nevertheless,<br />

in principle, <strong>TOMBO</strong> code can handle all atoms by increasing the cutoff energy<br />

for Plane Waves.<br />

11. Lines 12 to 18 are related to the radius of non-overlapping atomic sphere and the<br />

choice of AOs for each atom.<br />

12. In line 19, mesh of the unit cell division is defined if necessary. Otherwise default<br />

value is used for mesh. For accurate calculation, large mesh such as 128 or 192 is<br />

recommended as well as noz, which should be defined in INPUT.inp.<br />

An example of COORDINATES.inp for rutile TiO 2 crystal [26] is shown in Table 3.2.<br />

Table 3.2 COORDINATES.inp of rutile TiO 2<br />

line 1 system=Rutile_TiO2<br />

line 2 4.59<br />

line 3 1.000000000 0.000000000 0.000000000<br />

line 4 0.000000000 1.000000000 0.000000000<br />

line 5 0.000000000 0.000000000 0.643883326<br />

line 6 6<br />

line 7 Direct<br />

line 8 O 0.553188499 0.553188499 0.250000000<br />

line 9 O 0.946811501 0.946811501 0.250000000<br />

line 10 O 0.446811501 0.053188499 0.750000000<br />

line 11 O 0.053188499 0.446811501 0.750000000<br />

line 12 Ti 0.250000000 0.250000000 0.250000000<br />

line 13 Ti 0.750000000 0.750000000 0.750000000<br />

line 14 Ti_rct=0.8<br />

line 15 Ti_nc=3<br />

line 16 Ti_nv=4<br />

line 17 Ti_ns=3<br />

line 18 Ti_np=2<br />

line 19 mesh=64<br />

line 19 END_PARA


3.1 COORDINATES.inp 25<br />

Fig. 3.1 Atomic orbitals (AOs) available in <strong>TOMBO</strong> code: s-, p-, and d-orbitals are implemented


3.2 INPUT.inp 26<br />

3.1.1 rct<br />

"XX_rct" is the radius of the non-overlapping atomic sphere (“rct”) of XX atom in units<br />

of [Å]. Here, note that user should write S _rct = 0.8 to define “rct” of S atom, for example,<br />

because the atomic symbols H, B, C, N, O, F, S, K, V, Y, and I are one character only,<br />

one blank is necessary between the atomic symbol and _. The sum of two “rct” values of<br />

adjacent atoms should not exceed the interatomic distance. The value of “rct” may strongly<br />

affect the one-shot GW calculation results. In such a case, “rct” should be chosen as small<br />

as possible, although the necessary cut-off energy for plane waves is increased.<br />

3.1.2 nc<br />

"XX_nc" is the number of core AOs for which no truncation is performed. One should<br />

write N _nc = 1 [default] to define “nc” of N atom (1s only), for example. For titanium, the<br />

core orbitals of Ti are 1s 2 2s 2 2p 6 , so that there are 3 core AOs: 1s, 2s and 2p; so Ti_nc = 3.<br />

3.1.3 nv<br />

"XX_nv" is the number of valence AOs for which truncation (to confine inside the atomic<br />

sphere) is performed. One should write N _nv = 2 [default] to define “nv” of N atom (2s<br />

and 2p), for example. For titanium, the valene orbitals of Ti are 3s 2 3p 6 3d 2 4s 2 , so that there<br />

are 4 valence AOs: 3s, 3p, 3d and 4s; so Ti_nv = 4.<br />

3.1.4 ns<br />

"XX_ns" is the number of s-type AOs for XX atom. One should write N _ns = 2 [default]<br />

to define “ns” of N atom (1s and 2s), for example. For titanium, 1s 2 2s 2 2p 6 3s 2 3p 6 AOs are<br />

used because 4s is quite extended. Therefore there are 3 s AOs: 1s, 2s and 3s; so Ti_ns = 3.<br />

3.1.5 np<br />

"XX_np" is the number of p-type AOs for XX atom. One should write N _np =<br />

1 [default] to define “np” of N atom (2p), for example. For titanium, AOs of Ti are<br />

1s 2 2s 2 2p 6 3s 2 3p 6 , so that there are 2 kinds of p orbits: 2p and 3p; so Ti_np = 2.<br />

3.2 INPUT.inp<br />

A typical form of INPUT.inp is shown in Table 3.3. # and mean comments.


3.2 INPUT.inp 27<br />

Table 3.3 Form of INPUT.inp<br />

line 1<br />

line 2<br />

line 3<br />

line 4<br />

line 5<br />

line 6<br />

line 7<br />

line 8<br />

line 9<br />

line 10<br />

line 11<br />

line 12<br />

line 13<br />

line 14<br />

line 15<br />

line 16<br />

line 17<br />

line 18<br />

line 19<br />

line 20<br />

line 21<br />

line 22<br />

line 23<br />

line 24<br />

line 25<br />

line 26<br />

line 27<br />

line 28<br />

line 29<br />

line 30<br />

line 31<br />

line 32<br />

line 33<br />

line 34<br />

name of the target system (arbitrary)<br />

iApp<br />

iAlg<br />

ippmG<br />

jmax<br />

deltaE<br />

npp<br />

nod<br />

nog<br />

ncut<br />

noz<br />

ncs<br />

nol<br />

nband<br />

q_eliminate<br />

icontinue<br />

isphcut<br />

iTotalE<br />

lpri<br />

nspin<br />

iMIX<br />

smixSCF<br />

nSDCG<br />

iSblp<br />

iSblp2<br />

smixTD<br />

iasym<br />

icalAO<br />

iCry<br />

precLDA<br />

iFTapp<br />

iCheb<br />

nstep<br />

END_PARA


3.2 INPUT.inp 28<br />

An example of INPUT.inp for one-shot GW calc. of rutile TiO 2 [26] is shown in Table 3.4.<br />

Table 3.4 INPUT.inp of rutile TiO 2<br />

line 1 system=Rutile_TiO2<br />

line 2 iApp =LGNN<br />

line 3 iAlg =D<br />

line 4 ippmG =4<br />

line 5 jmax =200<br />

line 6 deltaE =2.0d0<br />

line 7 npp =100<br />

line 8 nod =8<br />

line 9 nog =18<br />

line 10 ncut =5<br />

line 11 noz =18<br />

line 12 ncs =22<br />

line 13 nol =400<br />

line 14 nband =10<br />

line 15 q_eliminate =0.005d0<br />

line 16 icontinue =0<br />

line 17 isphcut =0<br />

line 18 iTotalE =0<br />

line 19 lpri =1<br />

line 20 nspin =0<br />

line 21 iMIX =0<br />

line 22 smixSCF =0.80d0<br />

line 23 nSDCG =8<br />

line 24 iSblp =80<br />

line 25 iSblp2 =15<br />

line 26 smixTD =0.80d0<br />

line 27 iasym =0<br />

line 28 icalAO =0<br />

line 29 iCry =1<br />

line 30 precLDA =0.0001d0<br />

line 31 iFTapp =0<br />

line 32 iCheb =0<br />

line 33 nstep =1<br />

line 34 END_PARA


3.2 INPUT.inp 29<br />

3.2.1 iApp<br />

G for GW, B for Bethe–Salpeter, D for both of them.<br />

1 st character indicates the method of the SCF Loop:<br />

L or N (None): the default, LDA.<br />

H: Hartree–Fock (available only for Γ-point calculations).<br />

G: Self-consistent GW (available only for Γ-point calculations).<br />

2 nd to 4 th characters indicate the method after the SCF Loop:<br />

G: one-shot GW calculation.<br />

B: Bethe–Salpeter calculation.<br />

D: one-shot GW + Bethe-Salpeter calculation.<br />

T: T-matrix calculation.<br />

S: Second-order Møller–Plesset calculation.<br />

M: molecular dynamics (MD) using Newton’s equation of motion for nuclei available<br />

for LDA only.<br />

X: structural relaxation using Broyden algorithm available for LDA only.<br />

A: Adiabaic LDA, electron dynamics calculation by solving the TD Kohn–Sham<br />

equation.<br />

W: Wave functions (WaveF_HOMO, WaveF_LUMO, ...) output.<br />

3.2.2 iAlg<br />

D: Matrix Diagonalization (recommended for small target systems)<br />

S: Steepest Descent<br />

C: Conjugate Gradient + RMM-DIIS + Davidson.<br />

3.2.3 iexc<br />

Type of the exchange-correlation functional.<br />

iexc =3 [default] means that Perdew–Zunger’s interpolation formula [27] is used for the<br />

exchange-correlation functional of the DFT. Usually, this is not written in INPUT.inp.<br />

3.2.4 iCHG<br />

Flag to output total Charge Density (ChargeDensity files) after the SCF loop is converged.<br />

Either 1 [on] or 0 [off] should be set if necessary. Default value of iCHG is 0.


3.2 INPUT.inp 30<br />

3.2.5 Ulevel, Llevel<br />

Levels to draw wave functions.<br />

When “W” is assigned in iApp to output wave functions, user can set "Ulevel = ***"<br />

and "Llevel = ***" to control the number of wave functions to be generated. "Ulevel" and<br />

"Llevel" indicate the topmost and lowermost levels, and all the wave functions between (or<br />

including) these levels will be generated. Default is that Llevel = noc-1 (HOMO-1) and<br />

Ulevel = noc+1 (LUMO+1). If the system is spin polarized, wave functions are generated<br />

for up and down spins separately.<br />

3.2.6 ippmG<br />

Selection of plasmon pole model or full ω integration.<br />

ippmG = 0 [default] (or 3) means the use of the generalized plasmon-pole (GPP) model<br />

in the evaluation of the correlation part of the self-energy, Σ c , while ippmG = 1 and ippmG<br />

= 2 mean the use of the plasmon pole model by Engel–Farid and von der Linden–Horsch.<br />

Note: ippmG = 1 (Engel–Farid) is not available for crystal system. If ippmG = 4 is set, the<br />

correlation part of self-energy, Σ c , is evaluated by using the full ω integration along the real<br />

axis and then turned 90 ◦ parallel to the imaginary axis (only positive part). If ippmG = 5 is<br />

set, the full ω integration along positive real axis is performed.<br />

3.2.7 jmax<br />

Number of mesh points used for the full ω integration.<br />

Note: Typically "jmax = 200" is used for ippmG = 4 and "jmax = 1000" for ippmG =<br />

5, but "jmax" should be set as 0 [default] for ippmG = 0-3 when plasmon pole models are<br />

used.<br />

If jmax = 200 is used with ippmG = 4, 201 points are set at 0.1 + 0.2n [eV] and 20.1 +<br />

(1 + 2n)i [eV] for n = 0, 100 along the positive real axis and then rotated 90 ◦ parallel to<br />

the positive imaginary axis. If jmax = 1000 is used with ippmG = 5, 1001 points are set at<br />

0.1 + 9.2n [eV].<br />

3.2.8 deltaE<br />

Finite difference ∆E in the numerical derivative of the renormalization factor Z.<br />

For the renormalization factor Z in Eq. (2.6), it is necessary to evaluate the derivative of<br />

the self-energy Σ(ε) with respect to ε. "deltaE" is the fine difference (in units of [eV]) of<br />

∆ε of this numerical derivative ∂Σ(ε)/∂ε| ε<br />

LDA. Default value is deltaE = 0.5d0 [eV].<br />

nk


3.2 INPUT.inp 31<br />

Note: When numerical ω integration is performed (ippmG=4,5), deltaE = 2.0 is recommended<br />

instead of the default value.<br />

3.2.9 npp<br />

Number of plasmon poles. Default value for npp is 999.<br />

Note: Typically 100 is chosen in the case of ippmG = 1 and 2. This parameter is ignored<br />

in the case of ippmG = 0, 3, 4, 5.<br />

3.2.10 nod<br />

Maximum number of nodes in PWs in each direction (necessary item for any type of<br />

calculation).<br />

Max number of |n 1 |, |n 2 |, |n 3 | of reciprocal lattice vector G = n 1 b 1 + n 2 b 2 + n 3 b 3 for<br />

PWs. Default value for "nod" is 10.<br />

The cut-off energy can be obtained by:<br />

E cut−o f f =<br />

( 2π × nod<br />

A 1<br />

) 2<br />

(3.2)<br />

in the unit of [Ry]. Here A 1 is the lattice constant (length of lattice vector a 1 ) in the unit of<br />

Bohr radius.<br />

Note: The cut-off energy can also be found in the "OUPUT.out" file.<br />

Note: Instead of giving nod, one may give Ecutoff, which is the cutoff energy of PWs,<br />

corresponding the maximum kinetic energy of PWs, in units of [Ry]. (For example, one may<br />

set Ecutoff = 24.d0 in INPUT.inp.) For isolated systems, 18 Ry is required for hydrogen<br />

atoms and 24 Ry required for transition metal, but 7 Ry is enough for carbon atoms and 3<br />

Ry is enough for sodium and lithium atoms. In principle, one has to check the convergence<br />

of the total energy or level energies with increasing the cut-off energy (or "nod"). For crystal<br />

systems, because one has to reduce the radii of non-overlapping atomic spheres to, e.g., 0.8<br />

[Å], "nod" (or "Ecutoff") should be chosen larger.<br />

3.2.11 nog<br />

Maximum number of nodes in G waves in the Fock exchange (necessary item for HF<br />

and GW calculations only.)<br />

Max number of |n 1 |, |n 2 |, |n 3 | of reciprocal lattice vector G = n 1 b 1 + n 2 b 2 + n 3 b 3 for<br />

the exchange part of the self-energy Eq. (2.31). “nog” is also used for the Fourier space


3.2 INPUT.inp 32<br />

calculation of the asymmetric potential inside the non-overlapping atomic sphere if iasym =<br />

1 is set. Default value for "nog" is 20.<br />

Note: Typically set double of nod, because the spatially localized core contribution is<br />

quite important in the Fock exchange term.<br />

Note: The cut-off energy can also be found in the "OUPUT.out" file.<br />

3.2.12 ncut<br />

Maximum number of nodes in G,G’ waves in the polarization function and the correlation<br />

part of the self-energy (necessary item for GW calculation only).<br />

Max number of |n 1 |, |n 2 |, |n 3 | of reciprocal lattice vector G = n 1 b 1 +n 2 b 2 +n 3 b 3 for the<br />

polarization function and the correlation part of the self-energy Eq. (2.33). Default value<br />

for "ncut" is 8.<br />

Note: Typically set the same number as nod or a little less than nod. (necessary item for<br />

GW calculations).<br />

Note: The cut-off energy can also be found in the "OUPUT.out" file.<br />

3.2.13 noz<br />

Number of Fourier mesh for nuclear Coulomb tail.<br />

A smooth polynomial function inside the non-overlapping sphere connected to the 1/r<br />

long tail of nuclear Coulomb potential (see Fig. 1.4) is analytically Fourier transformed as<br />

Eq. (1.29) and summed over all atoms. "noz" is the maximum number of nodes in G waves<br />

in this Fourier transformation. It is then numerically Fourier back transformed along the<br />

radial direction in each atomic sphere to make spherical potential felt by AOs as Eq. (1.30).<br />

Default is noz = 24. For accurate calculation, large mesh such as 128 or 192 is recommended<br />

as well as mesh, which should be defined in COORDINATES.inp.<br />

3.2.14 ncs<br />

Number of core states.<br />

"ncs" is the number of core states, which is excluded in the calculation of the correlation<br />

part of the self-energy in the GW calculation or fixed in the TDDFT dynamics in the<br />

adiabatic LDA calculation. The default value is ncs = 0.<br />

For example, in rutile TiO 2 , there are 2 Ti atoms and 4 O atoms in a unit cell:<br />

Ti: 1s 2 2s 2 2p 6 3s 2 3p 6 1 + 1 + 3 +1 +3 = 9<br />

Note: p orbital contains p x , p y and p z , so the number is 3 for p orbital.


3.2 INPUT.inp 33<br />

O : 1s 2<br />

There is only one s orbital, so the number is 1 for s orbital.<br />

In rutile TiO 2 , 9 × 2 + 1 × 4 = 22, sothat ncs = 22.<br />

3.2.15 ncc<br />

Number of core constituents.<br />

"ncc" is the number of core constituents to be excluded from the Fock exchange in HF<br />

and GW calculations. For the intermediate states in the calculation of the exchange term,<br />

the states from “number of core constituents” (ncc) + 1 to “number of occupied states” (noc)<br />

are used in the code. Deep core states also contribute largely to the exchange term, so that<br />

"ncc" should be usually set at 0 [default].<br />

3.2.16 nol<br />

Number of levels.<br />

"nol" is the number of levels (states) to be included in the GW calculations of the correlation<br />

parts (the polarization function and the self-energy). Default value for iSblp is 300.<br />

Note: If "nol" is not enough, it leads band gap become large.<br />

Note: "nol" must be less than the number of basis set (functions) "nbs". The value of<br />

"nbs" can be found in "OUTPUT.out" file.<br />

3.2.17 nband<br />

Number of empty states to display in GWA.out.<br />

"nband" is the number of output empty levels (band), i.e., the number of empty states in<br />

conduction bands to be displayed in GW (GWA.out). Default is nband = 0.<br />

3.2.18 q_eliminate<br />

Criterion to remove AO.<br />

AO is eliminated, if truncated AO has norm less that "q_eliminate". For example, if<br />

one sets q_eliminate = 0.001 as small enough, no AO will be removed from the basis set.<br />

The default value of q_eliminate is 0.3, if the unit cell size is less than 4.0 Å, and 1 × 10 −2<br />

otherwise.<br />

3.2.19 q_min_mod<br />

Criterion to truncate AOs.


3.2 INPUT.inp 34<br />

Whether the subtraction of a polynomial from AO is performed or not depends on the<br />

norm of the original AO outside the atomic sphere. If this norm exceeds q_min_mod,<br />

the subtraction is performed. q_min_mod can be set in INPUT.inp. The default value of<br />

q_min_mod is 1 × 10 −4 .<br />

3.2.20 icontinue<br />

continue job to TDDFT dynamics or to one-shot GW.<br />

Frag to continue to the previous calculation by skipping the LDA SCF Loop. Set 0<br />

[default] to perform a new calculation or 1 to skip the LDA SCF Loop.<br />

3.2.21 isphcut<br />

Coulomb spherical cut (available for isolated systems only).<br />

-1 : Flag to truncate Coulomb potential tail at half of the lattice constant;<br />

0 : (default) not to truncate.<br />

3.2.22 iTotalE<br />

Total Energy calculation.<br />

To calculate SCF (LDA, HF, or self-consistent GW) Total Energy, set iTotalE = 1, but<br />

otherwise set iTotalE = 0 [default].<br />

3.2.23 lpri<br />

Output detailed information of calculations in ISYS.out.<br />

Either 1 [on] or 0 [off] should be set. Default value of lpri is 0.<br />

3.2.24 nspin<br />

Number up spin electrons minus number of down spin electrons.<br />

For the spin polarized system, one can set nspin = 1 for spin doublet. nspin = 2 for spin<br />

triplet, ... nspin = # for spin (# + 1) multiplet in INPUT.inp. Its default value is nspin = 0. In<br />

order to treat an antiferromagnetic system having zero total magnetic moment, one should<br />

write nspin = -1 in INOUT.inp and simultaneously, one may write the spin magnetic moment<br />

of each atom as m 1 or m -1 at the right of the atomic coordinate in COORDINATES.inp.<br />

(nspin = -1 can be used together with iExcite = 1, 2, ...)


3.2 INPUT.inp 35<br />

3.2.25 irelative<br />

Semi-relativistic effect.<br />

Flag to include the semi-relativistic effect into the LDA calculation (1 for ON, 0 for<br />

OFF). The Darwin and mass-velocity terms are included as a relativistic effect in the LDA<br />

calculation, when irelative = 1 is set.<br />

3.2.26 nion<br />

Number of excess electrons for ionic system.<br />

For isolated systems using spherical Coulomb cut-off (isphcut = -1), user can define<br />

“nion” to calculate ionic systems. Since nion means the excess number of electrons, so that<br />

nion > 0 corresponds to minus ion, and nion < 0 corresponds to plus ion. For example, nion<br />

= +2 corresponds to -2 ion. Note that plus and minus signs are different from the usual<br />

convention. Of course, nion = 0 (defual) means a neutral system.<br />

3.2.27 iExcite<br />

Number of levels to excite an up-spin electron.<br />

If one wants to excite one electron from HOMO to LUMO +n at the outset, one may<br />

define iExcite = 1 + n in INPUT.inp (n can be 0, 1, 2, ...). Default value of iExcite is 0,<br />

which means no excitation. Since only up-spin electron is excited, so that nspin should be<br />

set either -1 or some positive numbers.<br />

3.2.28 nonadiabatic<br />

Nonadiabatic treatment in TDDFT dynamics simulations.<br />

Nonadiabatic correction proportional to the velocity of atoms can be treated in the update<br />

of wave packet if nonadiabatic = 1 is set. Default value is nonadiabatic = 0.<br />

3.2.29 iMIX<br />

Method of SCF Charge Density Mixing.<br />

0: Linear, 1: Optimized Linear, 2: Pulay (DIIS), 3: Broyden Charge Mixing. Default<br />

setting is iMIX = 2. But, usually, iMIX = 3 is recommended in usual cases. Usually iMIX =<br />

3 (or 2) is recommended, but, if not converged, set iMIX = 0 with smixSCF = 0.90 or 0.95.


3.2 INPUT.inp 36<br />

3.2.30 smixSCF<br />

Charge Mixing Rate in SCF loop for LoopTD = 1.<br />

Rate of mixing (MAX) previous Charge Density for the 1 st MD step (LoopTD = 1).<br />

Default value is smixTD = 0.8d0.<br />

3.2.31 nSDCG<br />

Number of Steepest Descent Loops.<br />

Number of Steepest-Descent (SD) or Conjugate-Gradient (CG) iterations without updating<br />

the Hamiltonian. Default value is nSDCG = 8.<br />

3.2.32 iSblp<br />

Max SCF Loop for LoopTD = 1.<br />

Number of max SCF Loops for the 1 st step in the starting SCF Loop (LoopTD = 1).<br />

Default value for iSblp is 999.<br />

3.2.33 iSblp2<br />

Max SCF Loop with only Gamma point. Default value for "iSblp2" is 1.<br />

LoopSCF > iSblp2 : special point loop using SPOINT. Default value for "iSblp" is 1.<br />

3.2.34 mSblp<br />

Maximum number of SCF loops during MD or relaxation.<br />

Number of max SCF Loops in the dynamics Loop (LoopTD > 1). During this loop,<br />

Hamiltonian is updated with fixed atomic positions. Default value for "mSblp" is 1.<br />

In order to perform a simulation with conserved total energy, it is recommended to set<br />

mSblp = 20.<br />

3.2.35 smixTD<br />

Charge Mixing Rate (in or outside SCF loop) for LoopTD > 1.<br />

Rate of mixing (MAX) previous Charge Density after the 1 st MD step (LoopTD > 1).<br />

Default value is smixTD = 0.5d0.


3.2 INPUT.inp 37<br />

3.2.36 iasym<br />

Asymmetric potential inside atomic sphere.<br />

Frag to treat asymmetric AO charge inside atomic spheres (1 for ON, 0 for OFF). Default<br />

value of "iasym" is 0.<br />

3.2.37 icalAO<br />

AO redetermination at each SCF loop.<br />

Frag to update Atomic Orbitals (AOs) at each SCF step (1 for ON, 0 for OFF). Default<br />

value of "icalAO" is 0.<br />

3.2.38 iCry<br />

Cluster/Crystal Calculation.<br />

Frag to calculate crystal calculation (0 for cluster calculation, 1 for crystal calculation).<br />

Default set is iCry = 0, i.e., cluster calculation.<br />

3.2.39 precLDA<br />

Convergence criterion for Total Energy in units of [eV].<br />

Criterion of Total Energy convergence (SCF loop ends when ∆E becomes less than this).<br />

Default value is precLDA = 0.0001d0.<br />

3.2.40 iFTapp<br />

If iFTapp = 1 is set, an approximate treatment of this 1D (in radial coordinate inside<br />

atomic sphere) Fourier transformation is used to accelerate the computational speed. Default<br />

value of iFTapp is 0.<br />

3.2.41 iCheb<br />

Chebyshev fitting for atomic potential to accelerate the computational speed.<br />

Frag to fit AO potential in atomic spheres to Chebyshev polynomial (1 for ON, 0 for<br />

OFF), in order to reduce the time required for the computation of < PW|V |AO > matrix<br />

elements. Default value of iCheb is 1, i.e., Chebyshev fitting ON.


3.3 KPOINT.inp 38<br />

3.2.42 Mcheb<br />

Number of Chebyshev polynomials for fitting AO potential in atomic spheres. Default<br />

value of MCheb is 30.<br />

3.2.43 nStep<br />

Maximum number of MD (or Relaxation or TDDFT) steps. Default value of nStep is 1.<br />

3.2.44 singlet, triplet<br />

Initial guess for the singlet and triplet optical gaps.<br />

In solving the Bethe–Salpeter equation to obtain photoabsorption spectra, an initial<br />

guess for optical gap ("singlet" and "triplet") must be given in units of [eV]. "singlet" is<br />

a value for the singlet exciton, and "triplet" is a value for the triplet exciton. Default values<br />

are singlet = 0.d0 and triplet = 0.d0.<br />

3.3 KPOINT.inp<br />

In a crystal calculation (iCry = 1), user has to define k points (usually on symmetry<br />

lines) for the energy-band outputs. For each band, a list of energies of all the k points are<br />

given by "band_0001.out", "band_0002.out", .... The total list including all bands is given<br />

by band.out. Form of KPOINT.inp is given in Table 3.5.<br />

Table 3.5 Form of KPOINT.inp file<br />

line 1 Direct/Cartesian<br />

line 2 numbers of k points for "out+sum", "out", and "sum"<br />

line 3 1 st k point and weight<br />

line 4 2 nd k point and weight<br />

line 5 3 rd k point and weight<br />

line 6 ... ...<br />

line 7 n th k point and weight<br />

For a three-dimensional lattice defined by its primitive vectors (a 1 , a 2 , a 3 ), the reciprocal<br />

lattice is generated by its three reciprocal primitive defined by the formulae<br />

a 2 × a 3<br />

b 1 = 2π<br />

a 1 · (a 2 × a 3 ) ,<br />

b a 3 × a 1<br />

2 = 2π<br />

a 2 · (a 3 × a 1 ) ,<br />

b a 1 × a 2<br />

3 = 2π<br />

a 3 · (a 1 × a 2 ) . (3.3)


3.3 KPOINT.inp 39<br />

Note: line 1<br />

Direct: k points in the fractional coordinates (coefficients for b 1 , b 2 , b 3 .);<br />

Cartesian: coefficients for 2π/A 1 in Cartesian coordinates; (A 1 is the lattice constant.)<br />

Note: line 2<br />

For the LDA band calculation, “out + sum” should be equal to the number of k points<br />

listed below; “sum” and “out” should be set equal to 0. For the one-shot GW calculation,<br />

“out” is the number of k-points for output and “sum” is the number of k-points for the<br />

k-point sampling in the calculation of the polarization function for the correlation part of<br />

the self-energy (k points in “out” are inside the irreducible Brillouin Zone (BZ), but “sum”<br />

should be given in the whole BZ. Usually, “sum” should be given on Γ grid.); “out + sum”<br />

is the number of k-points for both “out” and “sum”.<br />

Note: line 3-7<br />

weights for “out+sum” and “sum” are meaningful to calculate the polarization function.<br />

Table 3.6 KPOINT.inp for LDA of Si<br />

line 1 Direct<br />

line 2 16 0 0<br />

line 3 0.500000000000 0.250000000000 0.750000000000 0.11764705882<br />

line 4 0.500000000000 0.333333333333 0.666666666667 0.05882352941<br />

line 5 0.500000000000 0.416666666667 0.583333333333 0.05882352941<br />

line 6 0.500000000000 0.500000000000 0.500000000000 0.05882352941<br />

line 7 0.375000000000 0.375000000000 0.375000000000 0.05882352941<br />

line 8 0.250000000000 0.250000000000 0.250000000000 0.05882352941<br />

line 9 0.125000000000 0.125000000000 0.125000000000 0.05882352941<br />

line 10 0.000000000000 0.000000000000 0.000000000000 0.05882352941<br />

line 11 0.100000000000 0.000000000000 0.100000000000 0.05882352941<br />

line 12 0.200000000000 0.000000000000 0.200000000000 0.05882352941<br />

line 13 0.300000000000 0.000000000000 0.300000000000 0.05882352941<br />

line 14 0.400000000000 0.000000000000 0.400000000000 0.05882352941<br />

line 15 0.500000000000 0.000000000000 0.500000000000 0.05882352941<br />

line 16 0.500000000000 0.125000000000 0.625000000000 0.05882352941<br />

line 17 0.437500000000 0.312500000000 0.750000000000 0.05882352941<br />

line 18 0.375000000000 0.375000000000 0.750000000000 0.05882352941<br />

Table 3.6 is KPOINT.inp for a LDA calculation of Si. The fourth column, i.e., the<br />

weight, is no meaning in the LDA calculation. The k points listed in Line 3, line 6, line


3.4 QPOINT.inp 40<br />

10, line 15, and line 18 corresponds to the W, L, Γ, X, and K points, respectively. The 1st<br />

Brillouin zone of bcc and fcc crystals is shown in Fig. 3.2.<br />

Fig. 3.2 The first Brillouin zone of bcc and fcc crystals.<br />

Table 3.7 is KPOINT.inp for a one-shot GW calculation of rutile TiO 2 [26].<br />

Table 3.7 KPOINT.inp for one-shot GW of rutile TiO 2<br />

line 1 Direct<br />

line 2 6 0 2<br />

line 3 0.00000 0.00000 0.00000 1.00 #Γ<br />

line 4 0.00000 0.00000 0.50000 1.00 #Z<br />

line 5 0.00000 0.50000 0.00000 1.00 #X<br />

line 6 0.00000 0.50000 0.50000 1.00 #R<br />

line 7 0.50000 0.50000 0.00000 1.00 #M<br />

line 8 0.50000 0.50000 0.50000 1.00 #A<br />

line 9 0.50000 0.00000 0.50000 1.00<br />

line 10 0.50000 0.00000 0.00000 1.00<br />

3.4 QPOINT.inp<br />

q points for the momentum transfer on Γ grid inside the irreducible BZ. This file is<br />

required for the one-shot GW crystal calculation only. Form of QPOINT.inp file is given by<br />

Table 3.8.


3.4 QPOINT.inp 41<br />

Table 3.8 Form of QPOINT.inp file<br />

line 1 Direct/Cartesian<br />

line 2 The number of q points in the irreducible zone, and the number of all grid points (see Appendix B<br />

in the whole zone<br />

line 3 1 st q point and weight<br />

line 4 2 nd q point and weight<br />

line 5 3 rd q point and weight<br />

line 6 ... ...<br />

line 7 n th q point and weight<br />

Note:<br />

The 3rd line should be "0.00000 0.00000 0.00000".<br />

Note:<br />

Increasing the number of q-points, it improves the accuracy of GW results, but it will<br />

take too much time.<br />

Note: line 1<br />

This line is the same as KPOINT.inp.<br />

Note: line 2<br />

Rutile TiO 2 has the symmetry group of P42/MNM, and 16 symmetry operations.<br />

Here, 6 q-points are chosen as shown in Fig. (3.3).<br />

Fig. 3.3 q-points and wights in b 1 , b 2 plane


3.5 SPOINT.inp 42<br />

Table 3.9 QPOINT.inp of rutile TiO 2<br />

line 1 Direct<br />

line 2 6 27<br />

line 3 0.00000 0.00000 0.00000 1.00<br />

line 4 0.00000 0.00000 0.50000 1.00<br />

line 5 0.50000 0.00000 0.00000 2.00<br />

line 6 0.50000 0.00000 0.50000 2.00<br />

line 7 0.50000 0.50000 0.00000 1.00<br />

line 8 0.50000 0.50000 0.50000 1.00<br />

3.5 SPOINT.inp<br />

In SPOINT.inp file, give special k points used for the SCF iteration loop in a crystal<br />

calculation. Usually, special points are on the Monkhorst–Pack grid inside the irreducible<br />

Brillouin zone. Form of SPOINT.inp file is given by Table 3.10.<br />

Table 3.10 Form of SPOINT.inp file<br />

line 1 Direct/Cartesian<br />

line 2 number of k points<br />

line 3 1 st k point and weight<br />

line 4 2 nd k point and weight<br />

line 5 3 rd k point and weight<br />

line 6 ... ...<br />

line 7 n th k point and weight<br />

Note: line 1<br />

This line is the same as KPOINT.inp.<br />

Table 3.11 is SPOINT.inp for a LDA calculation of Si. Here the fourth column, i.e., the<br />

weigh, has important meaning when the special point sampling is performed to make the<br />

total electron density.


3.5 SPOINT.inp 43<br />

Table 3.11 SPOINT.inp for LDA of Si<br />

line 1 Direct<br />

line 2 2<br />

line 3 0.2500000 0.2500000 0.2500000 1.00<br />

line 4 0.2500000 0.2500000 -0.2500000 3.00<br />

Table 3.12 is SPOINT.inp for a one-shot GW calculation<br />

Table 3.12 SPOINT.inp for one-shot GW of rutile TiO 2<br />

line 1 Direct<br />

line 2 24<br />

line 3 0.40000000 0.40000000 0.43750000 0.04000000<br />

line 4 0.40000000 0.40000000 0.31250000 0.04000000<br />

line 5 0.40000000 0.40000000 0.18750000 0.04000000<br />

line 6 0.40000000 0.40000000 0.06250000 0.04000000<br />

line 7 0.40000000 0.20000000 0.43750000 0.08000000<br />

line 8 0.40000000 0.20000000 0.31250000 0.08000000<br />

line 9 0.40000000 0.20000000 0.18750000 0.08000000<br />

line 10 0.40000000 0.20000000 0.06250000 0.08000000<br />

line 11 0.40000000 0.00000000 0.43750000 0.04000000<br />

line 12 0.40000000 0.00000000 0.31250000 0.04000000<br />

line 13 0.40000000 0.00000000 0.18750000 0.04000000<br />

line 14 0.40000000 0.00000000 0.06250000 0.04000000<br />

line 15 0.20000000 0.20000000 0.43750000 0.04000000<br />

line 16 0.20000000 0.20000000 0.31250000 0.04000000<br />

line 17 0.20000000 0.20000000 0.18750000 0.04000000<br />

line 18 0.20000000 0.20000000 0.06250000 0.04000000<br />

line 19 0.20000000 0.00000000 0.43750000 0.04000000<br />

line 20 0.20000000 0.00000000 0.31250000 0.04000000<br />

line 21 0.20000000 0.00000000 0.18750000 0.04000000<br />

line 22 0.20000000 0.00000000 0.06250000 0.04000000<br />

line 23 0.00000000 0.00000000 0.43750000 0.01000000<br />

line 24 0.00000000 0.00000000 0.31250000 0.01000000<br />

line 25 0.00000000 0.00000000 0.18750000 0.01000000<br />

line 26 0.00000000 0.00000000 0.06250000 0.01000000


Chapter 4<br />

Output Files<br />

Note:<br />

1. In every calculation, output files includes OUTPUT.out, Status.out, and ISYS.out.<br />

There appear also CHARGE.bin for the next job when icontinue = 1 is set.<br />

2. In the dynamics calculation when “M”, “X” or “A” is assigned in iApp of INPUT.inp,<br />

output files include also: posit.dat and eigen.dat as well as MD_Coordinates.xyz and<br />

MD_Coordinates.arc.<br />

3. In the crystal calculation, output files includes also: band.out, and so on.<br />

4. If “W” is assigned in iApp of INPUT.inp, output files includes also: WaveF_HOMO-<br />

1.cube, WaveF_HOMO-1.grd, WaveF_HOMO-1.vasp, and so on.<br />

5. If iCHG = 1 is set in INPUT.inp, output files includes also: ChargeDensity.cube,<br />

ChargeDensity.grd, and ChargeDensity.vasp.<br />

6. In the GW calculation, output files include also: GWA.out.<br />

7. In the Bethe–Salpeter equation (BSE) calculation, output files include also: PhotoAbsorptionSpectra.out.<br />

4.1 OUTPUT.out<br />

In OUTPUT.out file, you can find the number of occupied states "noc", the number of<br />

plane waves "nw", the number of atomic orbitals "nao", and the cut-off energies for plane<br />

waves, Fock exchange and correlation, and so on. This file appears just after the execution<br />

starts, so you can check these information quickly.


4.2 Status.out 45<br />

4.2 Status.out<br />

In Status.out, the job process is displayed during execution. One can display it by typing<br />

"tail -f Status.out".<br />

4.3 ISYS.out<br />

In ISYS.out file, you can find the detailed output during execution. First, the information<br />

of the unit cell, the result of Herman-Skillman’s atomic LDA calculation for generating<br />

AOs, the symmetry operations, the information of the truncation of AOs are written in this<br />

ISYS.out. After the mixed-basis calculation starts, for example, the LDA eigenvalues at<br />

each SCF step and the force contributions at each TD step are written in this ISYS.out. In<br />

the GW calculations, some information of the polarization and dielectric functions are also<br />

written.<br />

4.4 posit.dat<br />

When MD or other dynamics calculation is performed, atomic positions and forces at<br />

each time step are written in position.dat.<br />

4.5 eigen.dat<br />

When MD or other dynamics calculation is performed, eigenvalues at each TD step are<br />

written in eigen.dat.<br />

4.6 MD_Coordinates.xyz, MD_Coordinates.arc<br />

When MD or other dynamics calculation is performed, atomic trajectories are written<br />

in these files for visualization. MD_Coordinates.xyz is for VMD, MD_Coordinates.arc for<br />

Materials Studio.<br />

4.7 band.out<br />

In a crystal calculation when iCry = 1 set in INPUT.inp, band.out is the result of energy<br />

eigenvalues (at sequential k points) in all bands. This file can be read, for example, by


4.8 WaveF_HOMO-1.cube, WaveF_HOMO-1.grd,<br />

WaveF_HOMO-1.vasp, and so on 46<br />

Microsoft EXCEL to draw the energy band diagram as a line graph.<br />

band_0001.out is the result of energy eigenvalues (at sequential k points) in the 1 st band;<br />

band_0002.out is the result of energy eigenvalues (at sequential k points) in the 2 nd band;<br />

band_0003.out is the result of energy eigenvalues (at sequential k points) in the 3 rd band;<br />

... ...<br />

4.8 WaveF_HOMO-1.cube, WaveF_HOMO-1.grd,<br />

WaveF_HOMO-1.vasp, and so on<br />

If “W” is assigned in iApp of INPUT.inp, wave output files are generated for all wave<br />

functions between Llevel and Ulevel assigned in INPUT.inp. Files with an extension “cube”<br />

is for GaussView, files with an extension “grd” is for Materials Studio, and files with an<br />

extension “vasp” is for VESTA.<br />

4.9 GWA.out<br />

"GWA.out" file saves the GW calculation results.<br />

In a GW cluster calculation, GWA.out looks like Table 4.1.<br />

In a GW crystal calculation, please find the sentence The Result of q-point sum in<br />

GWA.out file. Information above this sentence is for intermediate calculations and not the<br />

final result. The final result is below this sentence, and looks like Table 4.2 for rutile TiO 2 .<br />

Note: Loopk_k = n (n=1, 2, ...)<br />

n means n th k point, corresponding to KPOINT.inp file.<br />

Note: level with * symbol<br />

It means this level is the highest occupied level, and level *+1 is the lowest empty level.<br />

Note: exc(LDA) is the LDA exchange-correlation potential, Vxc<br />

LDA , in [eV].<br />

Note: eps(LDA) is LDA eigenvalue in [eV].<br />

Note: xg(Fock) is exchange part of self-energy, Σ x , in [eV].<br />

Note: slf(GWA) is correlation part of self-energy, Σ x , in [eV].<br />

Note: QP(GWA) is quasi-particle energy εn,k 0 in [eV].<br />

∫<br />

εn,k 0 = εLDA n,k +<br />

drdr ′ ψ LDA<br />

n,k<br />

∗<br />

(r)[Σ(r,r ′ ;ε LDA<br />

n,k<br />

LDA<br />

) −Vxc<br />

δ(r − r ′ )]ψn,k LDA (r′ ) (4.1)


4.9 GWA.out 47<br />

Note: RQP(GWA) is renormalized quasi-particle energy ε GW<br />

n,k<br />

ε GW<br />

n,k<br />

= ε0 n,k + (ε0 n,k − εLDA<br />

n,k )∂ΣGW n,k (ε)<br />

∂ε<br />

∣<br />

∣<br />

ε=ε<br />

LDA<br />

n,k<br />

in [eV],<br />

which is equal to Eq. (2.6). Here Z n,k is a renormalization factor defined by<br />

Z n,k =<br />

[<br />

1 − ∂ΣGW<br />

∂ε<br />

n,k (ε)<br />

] −1<br />

ε=ε LDA<br />

n,k<br />

Z n,k , (4.2)<br />

. (4.3)<br />

Table 4.1 GWA.out of Li 2 GW cluster calculation<br />

Loopk_q = 0 Loopk_k = 1<br />

Na_ 2: mesh= 48 size= 2.117 nod=6 nog=12 nol= 300 ncut= 6<br />

q=0 only, iDiag 0 Cor Renor<br />

LDA -relation -malized<br />

< xc > eigenvalue < x > < c > QP energy QP energy<br />

Na 2 exc(LDA) eps(LDA) xg(Fock) slf(GWA) QP(GWA) RQP(GWA)<br />

* 3 -5.5682 -3.5408 -7.3972 -0.8465 -6.2163 -5.6038<br />

4 -3.5836 -1.7990 -1.3020 -0.9539 -0.4713 -0.7052<br />

5 -3.8425 -1.5722 -1.2141 -1.0514 0.0048 -0.2900<br />

6 -3.8425 -1.5722 -1.2141 -1.0513 0.0049 -0.2899<br />

7 -3.0842 -1.0258 -0.9242 -1.0423 0.0920 -0.1084<br />

8 -1.7537 -0.0763 -0.2596 -0.8076 0.6102 0.4784<br />

9 -2.0835 -0.0312 -0.2372 -1.0481 0.7669 0.5806<br />

10 -2.0835 -0.0312 -0.2372 -1.0482 0.7669 0.5806<br />

11 -1.6420 0.4796 -0.1403 -0.8559 1.1254 0.9919<br />

12 -1.6420 0.4796 -0.1403 -0.8559 1.1254 0.9919<br />

13 -1.0370 1.0156 -0.0488 -0.5180 1.4857 1.4327<br />

Eg(LDA): 1.742 Eg(GWA): 5.745<br />

IP(LDA): 3.541 IP(GWA): 6.216


4.9 GWA.out 48<br />

Table 4.2 GWA.out of TiO 2 GW crystal calculation<br />

The Result of q-point sum<br />

Loopk_k = 1<br />

Level exc(LDA) eps(LDA) xg(Fock) slf(GWA) QP(GWA) RQP(GWA)<br />

23 -20.6370 -8.7741 -28.3616 7.9457 -8.5530 -8.6157<br />

24 -21.8867 -7.5498 -29.4496 8.0295 -7.0832 -7.2316<br />

25 -22.5537 -6.8436 -29.9669 7.2448 -7.0120 -6.9951<br />

26 -22.5484 -6.7963 -29.9329 7.2063 -6.9746 -6.9580<br />

27 -18.7654 3.1389 -20.0824 3.4084 5.2303 4.7500<br />

28 -20.1376 3.4787 -20.8328 3.3223 6.1058 5.4468<br />

29 -16.8424 3.7836 -18.2006 3.6425 6.0679 5.5987<br />

30 -20.8289 4.9844 -21.5778 3.3271 7.5625 7.0369<br />

31 -20.8219 5.0327 -21.5454 3.2865 7.5957 7.0859<br />

32 -19.8849 6.2660 -21.1415 3.2785 8.2879 8.0060<br />

33 -19.9192 6.2689 -21.1534 3.2696 8.3042 8.0448<br />

34 -20.8307 6.8520 -22.0483 3.4441 9.0785 8.5782<br />

35 -21.9237 8.1268 -22.6500 2.7348 10.1353 9.8225<br />

36 -21.7953 8.3549 -22.3521 2.7431 10.5412 10.1170<br />

37 -21.8323 8.3580 -22.3833 2.7383 10.5453 10.1039<br />

* 38 -21.5760 8.7647 -22.2072 2.9078 11.0413 10.4756<br />

39 -22.3485 10.4435 -13.0584 -4.9269 14.8067 14.1228<br />

40 -21.9881 10.7806 -13.3086 -4.4607 14.9993 14.1706<br />

41 -22.7476 10.9637 -13.6819 -4.7350 15.2944 14.1623<br />

42 -23.2351 11.0461 -13.7080 -5.0459 15.5273 14.5298<br />

43 -23.2203 11.0883 -13.6749 -4.9537 15.6800 14.6532<br />

44 -25.2912 13.0470 -16.3438 -5.1293 16.8650 16.2493<br />

45 -24.9397 13.1399 -15.1763 -5.7544 17.1488 16.1741<br />

46 -26.0406 15.1183 -17.3744 -4.6174 19.1671 18.1623<br />

47 -26.0336 15.1694 -17.3581 -4.5614 19.2835 18.2629<br />

48 -26.7895 16.2643 -17.1368 -5.6359 20.2812 19.1626<br />

Loopk_k = 2<br />

Level exc(LDA) eps(LDA) xg(Fock) slf(GWA) QP(GWA) RQP(GWA)<br />

23 -22.0740 -7.3505 -29.6158 7.7835 -7.1088 -7.1427<br />

24 -22.1741 -7.3054 -29.6361 7.3740 -7.3934 -7.3899<br />

25 -22.0695 -7.3024 -29.5826 7.7010 -7.1144 -7.1340<br />

26 -22.1688 -7.2586 -29.6010 7.2227 -7.4681 -7.4680<br />

... ...


4.10 PhotoAbsorptionSpectra.out 49<br />

4.10 PhotoAbsorptionSpectra.out<br />

Photoabsorption spectra calculated by solving the Bethe–Salpeter equation (BSE) is<br />

written in PhotoAbsorptionSpectra.out as Table 4.3.<br />

BSE routine is available only for a cluster calculation.<br />

Note: line 1: singlet and triplet is the values assigned in INPUT.inp. If not assigned,<br />

0.000000000 (eV) is written.<br />

Note: line 2: multiplicity 0 corresponds to singlet and multiplicity 1 (must be 3 in usual<br />

definition) corresponds to triplet.<br />

Note: lines 3-: First column is the photoabsorption energy in units of [eV], second and<br />

third columns are the absorption coefficients for singlet and triplet excitons, respectively.<br />

Table 4.3 OUTPUT of TiO 2 GW calculation<br />

singlet = 0.000000000 (eV), triplet = 0.000000000 (eV)<br />

multiplicity = 0 1 <br />

0.000000000 0.002468937 0.185138621<br />

0.005440000 0.002480351 0.192629777<br />

0.010880000 0.002491845 0.200584522<br />

0.016320000 0.002503421 0.209041862<br />

0.021760000 0.002515078 0.218044985<br />

0.027200000 0.002526818 0.227641812<br />

0.032640000 0.002538641 0.237885633<br />

0.038080000 0.002550548 0.248835838<br />

0.043520000 0.002562540 0.260558778<br />

0.048960000 0.002574617 0.273128761<br />

0.054400000 0.002586781 0.286629224<br />

energy (eV) singlet exciton triplet exciton


Chapter 5<br />

Examples<br />

5.1 LDA calculation of isolated dimers<br />

Figs. 5.1(a) and (b) represent the force and total energy of nitrogen dimer (N 2 ) in the LDA<br />

calculations with and without the spherical cut of Coulomb potential. In both calculations,<br />

unit cell is simple cubic with the lattice constant of 12 [Å], "nspin=0", "N _rct=0.50", and<br />

"nod=12" corresponding to the PW cut-off energy of 11.0 [Ry]. The two results with and<br />

without the spherical cut are almost the same. There is a minimum of the total energy at<br />

distance 1.15 [Å], where the Force crosses zero.<br />

(a) with spherical cut<br />

(b) without spherical cut<br />

Fig. 5.1 Force and total energy of nitrogen dimer (N 2 ) versus bond length in a calculation<br />

with and without spherical cut.


5.1 LDA calculation of isolated dimers 51<br />

Figs. 5.2(a) and (b) represent the force and total energy of iron dimer (Fe 2 ) in the LSDA<br />

calculations with and without the spherical cut of Coulomb potential. In both calculations,<br />

unit cell is simple cubic, "nspin=6", and "Fe_rct=1.00". The lattice constant is 12 [Å] (6.8<br />

[Å]) and "nod=17" ("nod=10") corresponding to the PW cut-off energy of 22.2 [Ry] (23.9<br />

[Ry]) for the calculation with (without) the spherical cut. The two results with and without<br />

the spherical cut are slightly different. There is a minimum of the total energy at distance<br />

2.00 [Å] (1.97 [Å]), where the Force crosses zero.<br />

(a) with spherical cut<br />

(b) without spherical cut<br />

Fig. 5.2 Force and total energy of iron dimer (Fe 2 ) versus bond length in a calculation with<br />

and without spherical cut.<br />

Figs. 5.3(a) and (b) represent the force and total energy of iron dimer (Ni 2 ) in the LSDA<br />

calculations with and without the spherical cut of Coulomb potential. In both calculations,<br />

unit cell is simple cubic, "nspin=2", and "Ni_rct=1.00". The lattice constant is 12 [Å] (6.8<br />

[Å]) and "nod=17" ("nod=10") corresponding to the PW cut-off energy of 22.2 [Ry] (23.9<br />

[Ry]) for the calculation with (without) the spherical cut. The two results with and without<br />

the spherical cut are slightly different. There is a minimum of the total energy at distance<br />

2.03 [Å] (2.01 [Å]), where the Force crosses zero.


5.1 LDA calculation of isolated dimers 52<br />

(a) with spherical cut<br />

(b) without spherical cut<br />

Fig. 5.3 Force and total energy of nickel dimer (Ni 2 ) versus bond length in a calculation<br />

with and without spherical cut.<br />

The COORDINATES.inp and INPUT.inp files for iron dimer (Fe 2 ) calculation is shown<br />

in Tables 5.1 and 5.2.<br />

Table 5.1 COORDINATES.inp of Fe 2<br />

line 1 SYSTEM=Fe2<br />

line 2 12.000000000000<br />

line 3 1.000000000 0.000000000 0.000000000<br />

line 4 0.000000000 1.000000000 0.000000000<br />

line 5 0.000000000 0.000000000 1.000000000<br />

line 6 2<br />

line 7 Cartesian<br />

line 8 Fe 0.0000000000 0.0000000000 0.0000000000<br />

line 9 Fe 1.1547005384 1.1547005384 1.1547005384<br />

line 10 Fe_rct=1.00<br />

line 11 Fe_ns=3<br />

line 12 Fe_nc=3<br />

line 13 Fe_nv=4<br />

line 14 mesh=64<br />

line 14 END_PARA


5.2 Molecular Dynamics 53<br />

Table 5.2 INPUT.inp of Fe 2<br />

line 1 SYSTEM=Fe2<br />

line 2 iApp =LMNN<br />

line 3 iAlg =D<br />

line 4 nod =14<br />

line 5 ncs =0<br />

line 6 noz =64<br />

line 15 icontinue =0<br />

line 16 isphcut =-1<br />

line 17 iTotalE =1<br />

line 18 lpri =1<br />

line 19 nspin =6<br />

line 20 iMIX =0<br />

line 21 smixSCF =0.90<br />

line 23 iSblp =300<br />

line 26 iasym =0<br />

line 27 icalAO =0<br />

line 28 iCry =0<br />

line 29 iExcite =0<br />

line 30 precLDA =0.000001d0<br />

line 31 iFTapp =0<br />

line 32 iCheb =0<br />

line 33 nstep =1<br />

line 34 END_PARA<br />

5.2 Molecular Dynamics<br />

As an example of a first-principles molecular dynamics simulation, Fig. 5.4 shows several<br />

snapshots of a chemical reaction producing a formic acid (HCOOH) molecule from a carbon<br />

dioxide (CO 2 ) molecule and two hydrogen atoms (2H) [2].<br />

In this MD simulation, COORDINATES.inp file given in Table 5.3 and INPUT.inp files<br />

given in Table 5.4 are used.


5.2 Molecular Dynamics 54<br />

Fig. 5.4 Snap shots of a molecular dynamics (MD) simulation producing HCOOH from<br />

CO 2 + 2H.<br />

Table 5.3 COORDINATES.inp of 2H + CO 2<br />

line 1 CO2+H2=>HCOOH<br />

line 2 8.000000000000<br />

line 3 1.000000000 0.000000000 0.000000000<br />

line 4 0.000000000 1.000000000 0.000000000<br />

line 5 0.000000000 0.000000000 1.000000000<br />

line 6 2<br />

line 7 Cartesian<br />

line 8 C 1.2000000000 0.0000000000 0.0000000000<br />

line 9 O 0.0000000000 0.0000000000 0.0000000000<br />

line 10 O 2.4000000000 0.0000000000 0.0000000000<br />

line 11 H 1.2000000000 1.7000000000 0.0000000000<br />

line 12 H 2.4000000000 -1.7000000000 0.0000000000<br />

line 13 C _rct=0.60<br />

line 14 C _ns=2<br />

line 15 C _np=1<br />

line 16 O _rct=0.50<br />

line 17 O _ns=2<br />

line 18 O _np=1<br />

line 19 mesh=48<br />

line 20 dTime=0.1d0<br />

line 21 END_PARA


5.3 GW calculation of rutile TiO 2 55<br />

Table 5.4 INPUT.inp of 2H + CO 2<br />

line 1 CO2+H2=>HCOOH<br />

line 2 iApp =LMNN<br />

line 3 iAlg =S<br />

line 4 nod =12<br />

line 5 ncs =0<br />

line 6 noz =48<br />

line 15 icontinue =0<br />

line 16 isphcut =0<br />

line 17 iTotalE =1<br />

line 18 lpri =1<br />

line 19 nspin =0<br />

line 20 iMIX =3<br />

line 21 smixSCF =0.80<br />

line 23 iSblp =100<br />

line 26 iasym =0<br />

line 27 icalAO =0<br />

line 28 iCry =0<br />

line 29 iExcite =0<br />

line 30 precLDA =0.00001d0<br />

line 31 iFTapp =0<br />

line 32 iCheb =0<br />

line 33 nstep =1000<br />

line 34 END_PARA<br />

5.3 GW calculation of rutile TiO 2<br />

Five input files are needed for the GW crystal calculation<br />

COORDINATES.inp (Table 3.2)<br />

INPUT.inp (Table 3.4)<br />

KPOINT.inp (Table 3.6)<br />

QPOINT.inp (Table 3.8)<br />

SPOINT.inp (Table 3.10)<br />

in addition to a sh file for submitting job. Table 5.5 is an example of sh file.


5.4 LDA calculation of rutile TiO 2 56<br />

Table 5.5 A sh file for submitting job (run.sh)<br />

mkdir tmp<br />

date<br />

./home/program/main<br />

date<br />

unsetenv LDR_CNTRL<br />

Note: "/home/program/main" is the file path of <strong>TOMBO</strong> code.<br />

When the job is finished, the GW result is list in "GWA.out" file.<br />

5.4 LDA calculation of rutile TiO 2<br />

Method: Self-consistent field (SCF) loop with special-point sampling (SPOINT.inp) and<br />

band structure calculations at the k-points (KPOINT.inp) on symmetry lines are performed<br />

within standard local density approximation (LDA) of density functional theory (DFT).<br />

In the LDA calculation, it needs 4 input files:<br />

COORDINATES.inp<br />

INPUT.inp<br />

KPOINT.inp<br />

SPOINT.inp<br />

The COORDINATES.inp and SPOINT.inp for LDA calculation are the same with those<br />

for GW calculation. In INPUT.inp file, some parameters should be modified:<br />

—————————————-<br />

iApp = LGNN → LNNN<br />

—————————————-<br />

For example, KPOINT.inp for LDA calculation is list as follow:<br />

line 1 Direct<br />

line 2 51 0 0<br />

line 3 0.00000000000 0.00000000000 0.0000000000 1.00 #Γ<br />

line 4 0.00000000000 0.00000000000 0.0500000000 1.00<br />

line 5 0.00000000000 0.00000000000 0.1000000000 1.00<br />

line 6 0.00000000000 0.00000000000 0.1500000000 1.00


5.4 LDA calculation of rutile TiO 2 57<br />

line 7 0.00000000000 0.00000000000 0.2000000000 1.00<br />

line 8 0.00000000000 0.00000000000 0.2500000000 1.00<br />

line 9 0.00000000000 0.00000000000 0.3000000000 1.00<br />

line 10 0.00000000000 0.00000000000 0.3500000000 1.00<br />

line 11 0.00000000000 0.00000000000 0.4000000000 1.00<br />

line 12 0.00000000000 0.00000000000 0.4500000000 1.00<br />

line 13 0.00000000000 0.00000000000 0.5000000000 1.00 #Z<br />

line 14 0.00000000000 0.05000000000 0.4500000000 1.00<br />

line 15 0.00000000000 0.10000000000 0.4000000000 1.00<br />

line 16 0.00000000000 0.15000000000 0.3500000000 1.00<br />

line 17 0.00000000000 0.20000000000 0.3000000000 1.00<br />

line 18 0.00000000000 0.25000000000 0.2500000000 1.00<br />

line 19 0.00000000000 0.30000000000 0.2000000000 1.00<br />

line 20 0.00000000000 0.35000000000 0.1500000000 1.00<br />

line 21 0.00000000000 0.40000000000 0.1000000000 1.00<br />

line 22 0.00000000000 0.45000000000 0.0500000000 1.00<br />

line 23 0.00000000000 0.50000000000 0.0000000000 1.00 #X<br />

line 24 0.00000000000 0.50000000000 0.0500000000 1.00<br />

line 25 0.00000000000 0.50000000000 0.1000000000 1.00<br />

line 26 0.00000000000 0.50000000000 0.1500000000 1.00<br />

line 27 0.00000000000 0.50000000000 0.2000000000 1.00<br />

line 28 0.00000000000 0.50000000000 0.2500000000 1.00<br />

line 29 0.00000000000 0.50000000000 0.3000000000 1.00<br />

line 30 0.00000000000 0.50000000000 0.3500000000 1.00<br />

line 31 0.00000000000 0.50000000000 0.4000000000 1.00<br />

line 32 0.00000000000 0.50000000000 0.4500000000 1.00<br />

line 33 0.00000000000 0.50000000000 0.5000000000 1.00 #R<br />

line 34 0.05000000000 0.50000000000 0.4500000000 1.00<br />

line 35 0.10000000000 0.50000000000 0.4000000000 1.00<br />

line 36 0.15000000000 0.50000000000 0.3500000000 1.00<br />

line 37 0.20000000000 0.50000000000 0.3000000000 1.00<br />

line 38 0.25000000000 0.50000000000 0.2500000000 1.00<br />

line 39 0.30000000000 0.50000000000 0.2000000000 1.00<br />

line 40 0.35000000000 0.50000000000 0.1500000000 1.00<br />

line 41 0.40000000000 0.50000000000 0.1000000000 1.00<br />

line 42 0.45000000000 0.50000000000 0.0500000000 1.00


5.4 LDA calculation of rutile TiO 2 58<br />

line 43 0.50000000000 0.50000000000 0.0000000000 1.00 #M<br />

line 44 0.50000000000 0.50000000000 0.0500000000 1.00<br />

line 45 0.50000000000 0.50000000000 0.1000000000 1.00<br />

line 46 0.50000000000 0.50000000000 0.1500000000 1.00<br />

line 47 0.50000000000 0.50000000000 0.2000000000 1.00<br />

line 48 0.50000000000 0.50000000000 0.2500000000 1.00<br />

line 49 0.50000000000 0.50000000000 0.3000000000 1.00<br />

line 50 0.50000000000 0.50000000000 0.3500000000 1.00<br />

line 51 0.50000000000 0.50000000000 0.4000000000 1.00<br />

line 52 0.50000000000 0.50000000000 0.4500000000 1.00<br />

line 53 0.50000000000 0.50000000000 0.5000000000 1.00 #A<br />

Note:<br />

1 st k point, 2 nd k point, ..., and 51 st k point are listed from line 3 to line 53, and the<br />

energy-band outputs of these k points are "band_0001.out", "band_0002.out", ..., "band_0051.out".<br />

Note:<br />

In line 2, numbers of k points (the first number) is meaningful within LDA band structure<br />

calculation, but the second and third numbers( , 0, 0 ) is anyway required.<br />

Note:<br />

weight is meaningless in the LDA calculation and “out”-only k points, but should be<br />

given.)<br />

How to look at output data:<br />

band_0001.out energy eigenvalues (at sequential k points) in the 1 st band<br />

band_0002.out energy eigenvalues (at sequential k points) in the 2 nd band<br />

... ...<br />

band.out energy eigenvalues (at sequential k points) in all bands.<br />

With the GWA.out and band.out files, we obtain GW and LDA calculation results. We<br />

make the figure of band structure, shown in Fig. 5.5


5.5 Wave function calculation of rutile TiO 2 59<br />

Fig. 5.5 Band structure of the pure rutile TiO 2 . (Lines are the LDA results and dots are the<br />

GW results. The zero of energy is placed at the top of the valence band ([VBM] at the Γ<br />

point).[26]<br />

5.5 Wave function calculation of rutile TiO 2<br />

After GW calculation, we can continue Wave Function calculation. It needs to modify<br />

INPUT.inp file:<br />

—————————————-<br />

iApp = LGNN → LWNN<br />

icontinue = 0 → 1<br />

—————————————-<br />

and we make a new KPOINT.inp:<br />

Table 5.7 KPOINT.inp for Wave Function calculation<br />

line 1 Direct<br />

line 2 1 0 0<br />

line 3 0.00000 0.00000 0.00000 1.00 #Γ<br />

Note:<br />

line 3: the k pint of wave function calculation.<br />

Note:<br />

line 3: weight is meaningless.


5.5 Wave function calculation of rutile TiO 2 60<br />

When wave function calculation is finished, the below output files are obtained:<br />

WaveF_HOMO-1.cube<br />

WaveF_HOMO-1.vasp<br />

WaveF_HOMO-1.grd<br />

WaveF_HOMO.cube<br />

WaveF_HOMO.vasp<br />

WaveF_HOMO.grd<br />

WaveF_LUMO.cube<br />

WaveF_LUMO.vasp<br />

WaveF_LUMO.grd<br />

WaveF_LUMO+1.cube<br />

WaveF_LUMO+1.vasp<br />

WaveF_LUMO+1.grd<br />

Note: "HOMO" is the highest occupied state, and "LUMO" is the lowest empty state.<br />

Here, wave functions are calculated at Γ point (0, 0.0, 0.0) and R point (0, 0.5, 0.5).<br />

From the results of *.cube files, contour plots of the partial charge densities on (0 0 1) plane<br />

are made with same magnification using VESTA software, shown in Fig. 5.6.<br />

(a) The highest occupied state at the Γ poin<br />

(b) The lowest empty state at the R point<br />

Fig. 5.6 Contour plot of the partial charge density of the highest occupied state and the<br />

lowest empty state of the pure rutile TiO 2 on (0 0 1) plane.[26]


References<br />

[1] K. Ohno, S. Ono, R. Kuwahara, Y. Noguchi, R. Sahara, Y. Kawazoe, and M. H. F.<br />

Sluiter. <strong>TOMBO</strong> Group. http://www.ohno.ynu.ac.jp/tombo/index.html.<br />

[2] S. Ono, Y. Noguchi, R. Sahara, Y. Kawazoe, and K. Ohno. Tombo: All-electron<br />

mixed-basis approach to condensed matter physics. Computer Phys. Comm., 189:20–<br />

30, 2015.<br />

[3] P. Hohenberg and W. Kohn. Inhomogeneous electron gas. Phys. Rev., 136(3B):B864–<br />

B871, 1964.<br />

[4] W. Kohn and L. J. Sham. Self-consistent equations including exchange and correlation<br />

effects. Phys. Rev., 140:A1133–A1138, 1965.<br />

[5] K. Ohno, F. Mauri, and S. G. Louie. Magnetic susceptibility of semiconductors by an<br />

all-electron first-principles approach. Phys. Rev. B, 56(3):1009–1012, 1997.<br />

[6] S.F. Boys and F. Bernardi. The calculation of small molecular interactions by the differences<br />

of separate total energies. some procedures with reduced errors. Mol. Phys.,<br />

19:553, 1970.<br />

[7] B. Liu and A.D. McLean. Accurate calculation of the attractive interaction of two<br />

ground state helium atoms. J. Chem. Phys., 59:4557, 1973.<br />

[8] K. Pachucki and J. Komasa. Gaussian basis sets with the cusp condition. Chem. Phys.<br />

Lett., 389:209, 2004.<br />

[9] T. Ohtsuki, K. Ohno, K. Shiga, Y. Kawazoe, Y. Maruyama, and K. Masumoto. Insertion<br />

of Xe and Kr atoms into C 60 and C 70 fullerenes and the formation of dimers. Phys.<br />

Rev. Lett., 81:967–970, Aug 1998.<br />

[10] K. Ohno, F. Mauri, and S. G. Louie. Magnetic susceptibility of semiconductors by an<br />

all-electron first-principles approach. Phys. Rev. B, 56:1009–1012, Jul 1997.<br />

[11] Y. Noguchi, S. Ishii, and K. Ohno. Two-electron distribution functions and short-range<br />

electron correlations of atoms and molecules by first principles T-matrix calculations.<br />

J. Chem. Phys., 125(11):114108, 2006.


References 62<br />

[12] M. S. Bahrany, M. H. F.. Sluiter, and Y. Kawazoe. First-principles calculations of hyperfine<br />

parameters with the all-electron mixed-basis method. Phys. Rev. B, 73:045111,<br />

2006.<br />

[13] E. Runge and E. K. U. Gross. Density-functional theory for time-dependent systems.<br />

Phys. Rev. Lett., 52(12):997, 1984.<br />

[14] U. Saalmann and R. Schmidt. Non-adiabatic quantum molecular dynamics: basic<br />

formalism and case study. Z. Phys. D, 38:153, 1996.<br />

[15] T. Sawada and K. Ohno. Time-dependent density functional approach to chemical<br />

reactions induced by electronic double excitations. Chem. Phys. Lett., 405:234, 2005.<br />

[16] R. W. Godby, M. Schlüter, and L. J. Sham. Self-energy operators and exchangecorrelation<br />

potentials in semiconductors. Phys. Rev. B, 37:10159–10175, 1988.<br />

[17] L. Hedin. New method for calculating the one-particle Green’s function with application<br />

to the electron-gas problem. Phys. Rev., 139(3A):A796–A823, 1965.<br />

[18] L. Hedin and S. Lundqvist. Effects of electron-electron and electron-phonon interactions<br />

on the one-electron states of solids. volume 23 of Solid State Physics, pages 1 –<br />

181. Academic Press, 1970.<br />

[19] M. S. Hybertsen and S. G. Louie. Electron correlation in semiconductors and insulators:<br />

Band gaps and quasiparticle energies. Phys. Rev. B, 34(8):5390–5413, 1986.<br />

[20] K. Ohno, K. Esfarjani, and Y. Kawazoe. Computational materials science: from ab<br />

initio to Monte Carlo methods, volume 129. Springer-Verlag, Berlin and Heidelberg,<br />

1999; see Fig. 2.23 on p.113.<br />

[21] M. S. Hybertsen and S. G. Louie. Theory of quasiparticle surface states in semiconductor<br />

surfaces. Phys. Rev. B, 38:4033–4044, Aug 1988.<br />

[22] R. W. Godby, M. Schlüter, and L. J. Sham. Accurate exchange-correlation potential for<br />

silicon and its discontinuity on addition of an electron. Phys. Rev. Lett., 56:2415–2418,<br />

Jun 1986.<br />

[23] R. W. Godby, M. Schlüter, and L. J. Sham. Quasiparticle energies in GaAs and AlAs.<br />

Phys. Rev. B, 35:4170–4171, Mar 1987.<br />

[24] R. W. Godby, M. Schlüter, and L. J. Sham. Trends in self-energy operators and their<br />

corresponding exchange-correlation potentials. Phys. Rev. B, 36:6497–6500, Oct 1987.<br />

[25] S. Ishii and K. Ohno. Full-ω integration GW calculation of Ne atom. unpublished.<br />

[26] M. Zhang, S. Ono, and K. Ohno. All-electron GW calculation of rutile tio 2 with and<br />

without nb impurities. Phys. Rev. B, 92(1):035205;1–7, 2015.<br />

[27] J. P. Perdew and Alex Zunger. Self-interaction correction to density-functional approximations<br />

for many-electron systems. Phys. Rev. B, 23:5048–5079, 1981.


Appendix A<br />

Materials Studio<br />

Materials Studio is software for simulating and modeling materials. It is developed<br />

and distributed by Accelrys. Now, I introduce how to use Materials Studio to build crystal<br />

model, 1 st Brillouin zone, and make SPOINT.inp file.<br />

Fig. A.1 Materials Studio


A.1 Build crystal model 64<br />

A.1 Build crystal model<br />

Step 1<br />

File → New... → 3D Atomistic<br />

Step 2<br />

Build → Crystals → Build crystal...<br />

Step 3<br />

Set "Space Group" and "Lattice Parameters".<br />

Fig. A.2 Build Crystal<br />

The default is space group P1, i.e. no symmetry. Rutile TiO 2 has the space group<br />

P42/MNM. By telling Materials Studio this symmetry it will automatically apply it to the<br />

atoms, thus generating atoms at the symmetry points.<br />

Add the lattice constant — click on the “Lattice” tab near the top of the “Build Crystal”<br />

window. Set Lengths and Angles.<br />

Step 4<br />

Build → Add Atoms


A.2 1 st Brillouin zone 65<br />

Fig. A.3 Add atoms<br />

Add atoms at the origin, by changing the “element” from its default and clicking “Add”.<br />

By default the co-ordinates are in fractionals, but you can change this on the “Options” tab.<br />

Note:<br />

For some common crystal model, it can be imported:<br />

File → Import... → Structures/Examples<br />

A.2 1 st Brillouin zone<br />

Step 1<br />

Modules → CASTEP → Calculation;<br />

Step 2<br />

Properties → Band structure → More...;<br />

Step 3<br />

choose "Path...";<br />

Step 4<br />

In "Brillouin Zone Path", click "Create", and choose "Display reciprocal lattice".


A.3 *.cell file 66<br />

Fig. A.4 1 st Brillouin zone<br />

A.3 *.cell file<br />

A.3.1<br />

make *.cell file<br />

Step 1<br />

Modules → CASTEP → Calculation;<br />

Step 2<br />

In "CASTEP Calculation", choose " k-points", and click "More...";


A.3 *.cell file 67<br />

Fig. A.5 CASTEP Calculation<br />

Step 3<br />

In "CASTEP Electronic Options", choose " Electronic", click "Custom grid parameters",<br />

and set a, b and c of k-points in "Grid parameters" using Monkhorst-Pack grid.<br />

Fig. A.6 CASTEP Electronic Options


A.3 *.cell file 68<br />

Step 4<br />

In "CASTEP Calculation", click "Files...", and click "Save Files". Then we can obtain<br />

the input files of Materials Studio.<br />

Step 5<br />

In "Project", right click, choose "Open Containing Folder".<br />

Fig. A.7 Input file 1<br />

Step 6<br />

Input files: (take ZnO for example) some files are hidden files.<br />

Fig. A.8 Input file 2<br />

A.3.2<br />

information of *.cell<br />

Take wurtzite ZnO for example, Open *.cell file, shown in Fig. A.9<br />

Note:


A.3 *.cell file 69<br />

Fig. A.9 ZnO.cell file<br />

From "%BLOCK LATTICE_CART" to "%ENDBLOCK LATTICE_CART", these are<br />

lattice paremeters in the Cartesian coordinates, can be used to make "CORRDINATE.inp"<br />

file;<br />

Note:<br />

From "%BLOCK KPOINTS_LIST" to "%ENDBLOCK KPOINTS_LIST", these are<br />

k-points for LDA/GGA calculation, can be used to make "SPOINT.inp" file.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!