• Aucun résultat trouvé

A force-constant model of graphene for conductivity calculations

N/A
N/A
Protected

Academic year: 2021

Partager "A force-constant model of graphene for conductivity calculations"

Copied!
1
0
0

Texte intégral

(1)

   

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1 30x30 smear 630K

60x60 smear 315K

       

       

       

   

   

A  force-­‐constant  model  of  graphene   for  conduc4vity  calcula4ons  

Henrique Miranda

(1)

and Ludger Wirtz

(1)

(1)

Physics  and  Materials  Science  Research  Unit,  University  of  Luxembourg,  Luxembourg  

Mo=va=on  

[1]  O.  Dubay  and  G.  Kresse,  Phys.  Rev.  B  67,  035401  (2003).  

[2]  T.  Sohier,  M.  Calandra,  C.-­‐H.  Park,  N.  Bonini,  N.  Marzari,  and  F.  Mauri,  Phys.  Rev.  B  90,  125414  (2014).  

[3]  J.  L.  Mañes,  Phys.  Rev.  B  76,  045430  (2007).  

[4]  R.  A.  Jishi,  L.  Venkataraman,  M.  S.  Dresselhaus,  and  G.  Dresselhaus,  Chemical  Physics  LeXers  209,  77  (1993).  

[5]  L.  Wirtz  and  A.  Rubio,  Solid  State  CommunicaZons  131,  141  (2004).  

[6]  Y.  N.  Gartstein,  Physics  LeXers  A  327,  83  (2004).  

[7]  S.  Piscanec,  M.  Lazzeri,  F.  Mauri,  A.  C.  Ferrari,  and  J.  Robertson,  Phys.  Rev.  LeX.  93,  185503  (2004).  

[8]  S.  Pisana,  M.  Lazzeri,  C.  Casiraghi,  K.  S.  Novoselov,  A.  K.  Geim,  A.  C.  Ferrari,  and  F.  Mauri,  Nat.  Mater.  6,  198  (2007).  

[8]  J.  Li,  H.  P.  C.  Miranda,  Y.  Niquet,  C.  Delerue,  L.  Genovese,  I.  Duchemin,  and  L.  Wirtz  (submiXed)    

 

Graphene  has  some  interes=ng  proper=es:  

•  high  charge-­‐carrier  mobility  (limited  by  electron-­‐phonon  interacZon)  

•  InteresZng  for  high  frequency  applicaZons  (control  of  mobility  is  important)   Aim  of  the  project:  

•  Create  a  4  nearest-­‐neighbor  (4NN)  force-­‐constant  model  of  graphene  that  reproduces  the   DFT  calculated  phonon  frequencies  and  modes.  

•  The  model  will  be  used  for  calculaZng  the  phonon-­‐limited  electrical  conducZvity  of  graphene   and  carbon  nanotubes.[8]  

•  Incorporate  long-­‐range  interacZons  in  the  model.  

Ab-­‐ini4o  phonon  dispersion  of  graphene  

Force  constant  model  

The force constants are second derivatives of the total energy with respect to the positions of two atoms:

They depend on:

•  Types of atoms interacting

•  Relative position vector

Transformation from local to global coordinates:

Build the dynamical matrix:

Calculate the phonon frequencies:

References  

Electron-­‐Phonon  coupling  

Comparison  of  4NN  models  

–  The off-diagonal elements of the force constant matrix are VERY important!

•  They can be modeled as angular springs between the carbon atoms [6].

•  4NN diagonal model enough to reproduce the dispersion but gives bad phonon eigenvectors

•  4NN model with 2 off-diagonal parameters: phonon eigenvectors and eigenvalues OK!

–  Simple model with few parameters (14 parameters instead of 12 ).

–  Strain dependent model by fitting to strained graphene.

–  Local to global coordinates transformation is delicate with off-diagonal terms.

–  The same analysis and model can be done for other materials (BN)

Conclusions  

•  Ground-­‐state:  LDA  norm-­‐conserving  pseudopotenZal,  35  Ha  plane-­‐wave  cut-­‐off,  4.65  Bohr  ladce   constant  (relaxed).  

•  Phonon  dispersion:  DFPT  sampling  of  30x30  q-­‐points.  

Cij = @2E

@xi@xj Global reference frame Cij0 = @2E

@x0i@x0j Local reference frame

C = R

T

C

0

R

   

Long-­‐range  force  constants  

D

ab

(~q) = X

s

C

sab

e

i~q.( ~ra r~s)

l/ti1 = ✏l/ti3 = 0

ti/l1 = ✏ti/l3 = 0

l/ti2 = ti/l2

l/ti4 = ✏ti/l4

C0n = 0 B@

lnl/tin 0

ti/ln tin 0

0 0 ton

1 CA

Off  diagonal  terms:  

ogen  not  included  [4]  

   

Phonon  displacements  

det |D

ab

(~q) !(~q)

2

| = 0

The  canonical  phonon  modes  are  defined  using  two  rules[2]:  

•  Eigenvector  of  a  longitudinal  (transverse)  mode  is  parallel  (perpendicular)  to  the  phonon’s  momentum  

•  Phase  differences  between  atoms  is                                                      for  acousZc  and                                                              for  opZcal  modes.  

 

The  DFT  calculated  phonon  modes  tend  to  the  canonical  ones  in  the  long  wavelength  limit.  

At  finite  momentum  there  is  an  acousZc/opZcal  mixing  linear  with  |q|[2-­‐3].    

               

Can  we  capture  this  dependence  correctly  in  a  4NN  model?  Why  is  it  important?  

•  Real  space  cut-­‐off  the  force  constants   up  to  4NN.  

•  Kohn  anomalies  are  suppressed  

•  Out-­‐of-­‐plane  mode  frequencies   become  imaginary  (unstable  system)  

•  The  4NN  model  does  not  reproduce   totally  the  phonon  dispersion.  

•  Can  we  add  an  analyZcal  correcZon?  

e

i~q.( ~r1 r~2)

e

i~q.( ~r1 r~2)

•  We  generate  the  dynamical  matrix  using  4NN  and  compare  to  the  ab-­‐ini&o  dynamical  matrix  and/or   the  eigenvalues  using  the  quality  funcZon  and  find  the  best  fit  using  a  simplex  method:  

   

•  We  calculate          from  the  phonon  eigenvectors  calculated  with  the  different  models:  

•  Good  agreement  with  reported  value  [2]      

eaq, ˜LA = 1

p2eiq·(R+ra) q

|q|

eaq, ˜T A = 1

p2eiq·(R+ra) q ?

|q ?|

eaq, ˜LO = a 1

p2eiq·(R+ra) q

|q|

eaq, ˜T O = a 1

p2eiq·(R+ra) q ?

|q ?|

TA   LA  

ZA   ZO  

TO  

LO   •  Kohn  anomalies  at  Γ  and  K  points  that  

connect  the  Fermi  surface  (non-­‐analyZc   phonon  dispersion).  

•  We  can’t  describe  the  full  dispersion   with  a  finite  set  of  force  constants  

(Fourier  interpolaZon  fails).  

•  The  electronic  temperature  changes   the  slope  of  the  phonon  dispersion.  

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

ab-initio ref. [4]

Diagonal  force-­‐constant  tensor  

l ti

to

xj xi

Phonon  dispersion  of  doped  graphene  

Phonon  dispersion  of  compressed/stretched  graphene  

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

ab-initio ref. [5]

Off-­‐diagonal  for  2NN  

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

ab-initio New fit

Off-­‐diagonal  for  2NN  and  4NN  

The  electron-­‐phonon  coupling  matrix  for  graphene  has  the  form[2-­‐3]:    

     

Where                and                are  related  to  the  acousZc-­‐opZcal  mixing  in  the  phonon  modes:  

   

Where              and              are  calculated  using  the  canonical  phonon  modes.  

To  get  the  correct  electron-­‐phonon  coupling  the  model  should  reproduce  the  correct  value  of            .  

•  Fit  4NN  force  constants  to  

strained  and  stretched  graphene.  

•  When  doping  graphene  we  

change  the  Fermi  energy  and  the   Kohn  anomaly  is  found  now  on  a   circle  in  the  Brillouin  zone  

 

CalculaZons  within  the  adiabaZc  approximaZon  

eq,LA =p

1 2|q|2eq, ˜LA |q|[sin(3✓q)eq, ˜LO + cos(3✓q)eq, ˜T O] eq,T A =p

1 2|q|2eq, ˜T A |q|[cos(3✓q)eq, ˜LO sin(3✓q)eq, ˜T O] eq,LO =p

1 2|q|2eq, ˜LO |q|[sin(3✓q)eq, ˜LA + cos(3✓q)eq, ˜T A] eq,T O =p

1 2|q|2eq, ˜T O |q|[cos(3✓q)eq, ˜LA sin(3✓q)eq, ˜T A]

Canonical  modes   Modes  with  Acous=c/Op=cal  mixing  

˜O

˜A

Acous=c  phonons   Op=cal  phonons  

A O

A

⇡ ˜

A

˜

O O

= ˜

O

n 1 2 3 4

ln 40.905 7.402 -1.643 -0.609

tin 16.685 -4.051 3.267 0.424

ton 9.616 -0.841 0.603 -0.501

l/tin 0.000 0.632 0.000 -1.092

ti/ln 0.000 -0.632 0.000 -1.092

[1]  

Model   Phonon  

frequencies   Phonon   displacements  

near  Γ  

Ref  [4]      

Ref  [5]      

New  Fit      

Table  2.  FiXed  force  constants  

-10 0 10 20 30 40 50 60 70

1 1.5 2 2.5 3 3.5 4

Force(dyn/cm)

Distance (˚A) long.

trans. in-plane trans. out-of-plane

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

M K

Energy(ev)

ef

AcousZc-­‐opZcal  mixing  

Table  1.  Comparison  of  the  different  models  

Ref.  [4]   Ref.  [5]   New  Fit  

Fermi energy

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

ef +.0168

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

nn 4

0 500 1000 1500 2000

M K

Frequencycm1

-5%

0 200 400 600 800 1000 1200 1400 1600

M K

Frequencycm1

5%

compressed   stretched  

Hqe-ph = i|q|

2↵(q)QLA Ae2i✓q(QLA + iQT A)

Ae 2i✓q(QLA iQT A) 2↵(q)QLA + i

0 Oei✓q(QLO + iQT O)

Oe i✓q(QLO iQT O) 0

Q = eq · eq,⌫

ProjecZon  on  the   canonical  modes  

⇡ 0.10˚ A

= ↵eig X

q=Nq

X

n

h!0(~q)n !(~q)abn i2

+ ↵dyn X

q=Nq

X

ij

X

a,b

hD0(~q)abij D(~q)abij i2

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

ref [4] ref [5]

New fit

ab-in

itio LD A

neighbor  

Références

Documents relatifs

In case of conservative force (field) potential energy is equal to negative of work done in shifting the body from reference position to given position.. This is why in shifting

Kreuzer, ‘Numerical calculation of listener-specific head- related transfer functions and sound localization: Microphone model and mesh discretization’,

Construct replicas of the unit cell with the size of the

In the research field of ionic conductivity in glasses, the idea of randomly varying ion jump activation energies has been around since the fifties.. The idea was not

Par conséquent, des stratégies devraient être élaborées dans la phase de la préparation pour restaurer les activités et infrastructures essentielles (comme les hôpitaux) le

R&um6 : Les modkles thermodynamiques classiques ne permettent pas une description rkaliste de la skgrkgation intergranulaire, car ils ne prennent pas en compte les

From the SEM images, we measured the average surface area of the CNS in this study (figures. We obtain the resistivity to be 150 µ Ωm in the reference sample and 3.5 µ Ωm in the

The ground state was now well below the Hartree-Fock result but, as expected, the wave functions contained dominant components lying outside our original truncated space.. Figure