LECTURE 7 : CALCULATING PROPERTIES Space correlation functions Radial distribution function Structure factor Coordination numbers, angles, etc.

Transcription

1 We have generated sets of (x,y,z) positions for various times at various given thermodynamic conditions (N,V,T,P). Now, we use them. LECTURE 7 : CALCULATING PROPERTIES Space correlation functions Radial distribution function Structure factor Coordination numbers, angles, etc.

2 A) RADIAL DISTRIBUTION FUNCTION Structural caracterization: Probability densities ρ N (1) and ρ N (2) i.e. one can find N particles and N(N-1) pairs of particles in the total volume, respectively. Note thattheseare n=1 and n=2 cases of a generalcorrelation function g (n) givenby : Z N the partition function, V the volume, N the nb of particles

3 So thatone can write for e.g. n=2: which definesthe radial distribution (rdf) g N2 (r 1,r 2 )function, alsogivenby For an homogeneousisotropic system, one has ρ N (1) =ρ. Dependence of the rdfonly on relative distance between particles: So that ρg(r) is the conditional probability to find another particle ata distance r away fromthe origin.

4 g(r) is the pair correlation function In a simulation box with PBC, one cannot obtain the structure beyond r > L/2. Alternatively: Radial distribution function : ()= 1 (+ ) Pair correlation function: ()= ( ) Alternative definitions : Total distribution function: =4() Differential distribution function =4 1

5 Pair distribution function : examples Visual inspection allows to distinguishbetweena crystalline and an amorphousstructure CCP Ni lattice with EAM potential

6 Effect of thermodynamic variables : temperature Amorphous Se The integral of g(r) allows to determine the number of neighbors around a central atom. Remember The integral to the first minimum gives the coordination number. r m = 4 Caprion, Schober, PRB 2000 Running coordination number N(r) ()= 4

7 Effect of thermodynamic variables : pressure d-geo 2 Direct comparison with experiments can fail Simple force fields can not account for pressure-induced changes (metallization) Additional structural insight isprovided by partial correlation functions : Ge-Ge, Ge-O, O-O Expt. (neutron) Salmon, JPCM 2011 MD: 256 GeO 2 using Oeffner-Elliot FF Micoulaut, JPCM 2004

8 Effect of composition: extending to multicomponent systems This is case for most glasses and materials SiO 2, GeSe 2, SiO 2 -Na 2 O, Consider a system with n components having N 1, N 2, N n particles. Wecan write $$ ()= ( ) $ with α between [1,n] And for! #: % % $& ()= ( $ ) & Out of which can be computed the pair correlation function: % ' = 1 ( ( (),( (can be also neutron or XRD weighted)

9 Ge 1 O 2 Pair distribution function in multicomponent systems: GeO 2 Evidence for neighbour correlations Ge 3 Ge 1 Ge 2 O 3 O 1 O 3 O 3 O 1 O 3 D. Marrocchelli et al., 2010 O 1 O 2

10 Pair distribution function in multicomponent systems: GeSe 2 Evidence for homopolar bonds Ge-Ge and Se-Se Prepeaks in relevant pdfs

11 Thermodynamic quantities from the computed g(r): Energy First, one needs to remember that the energy is related to the partition function Q(N,V,β) as: where : -.,#, = / 0 1! = 1 0 1! 3 4&5(6 ) 7 λ comesfromthe integration of the momentump in the phase spacegivenby: so that:

12 and In order to compute the average energy, one needs to compute the average potential U. Assume the case of a pairwise potential, so that: i.e. U isa sumof termsdepending only between2 particles, and thuscontains N(N-1) terms. Then, wecan write: (i.e. all terms in the first line are the exact same integrals, just with different labels)

13 which containsa two-body probability distribution function ρ (2) (r 1,r 2 ) sothat<u> can berewritten as : where weremind (slide 2) that: One thus arrives to the result:

14 Example : Glass transition in a model A-B glass Flores-Ruiz et al. PRB 2010 Reproduction of the glass transition phenomenon

15 Thermodynamic quantities from the computed g(r): Pressure We use the Maxwell relation: The volume dependence can be made explicit by changing the variables: sothatwehave fromthe partition function Z N the desiredderivativefromv: And: which involvesthe force F i acting on a particle i

16 Again, for, pairwiseinteractions, F i can be written as : and since one has (interchanging i-j summations): Using Newton s third law, we furthermore have : and the Ensemble average is given by:

17 Wedo the same as for the energy: and for a pair potential U pair : And finally: The pressure canbecomputedfromthe derivative of the pair potentialand fromg(r).

18 Examples: equation of state of liquids GeO 2 SiO 2 Shell et al. PRE 2002 Micoulaut et al. PRE 2006

19 B) STATIC STRUCTURE FACTOR For a simple liquid, the static structure factor isgivenby : where ρ k is the Fourier transformof the microscopic density ρ(r). This readsas: and, remembering that ρ one has: If isotropic and uniform medium, remember that ρ(r,r )=ρ 2 g(r,r )

20 Also, g(r,r ) depends only on r-r, i.e.: Or (isotropic fluid, everything depends only on k= k ): The calculation of the structure factor S(k) isachieved via a Fourier transformof the pair distribution function g(r).

21 Calculating a structure factor S(k) from a MD simulation : 2 options 1. Calculate the pair correlation function g(r) from the MD trajectory, then use : 2. Calculate directly S(k) from the trajectory using: As 2 Se 3 Differences between both methodscan arise, and one islimited to r<l/2, i.e. k<π/l Effects of the components of the wavevector

22 Direct comparison with experiments : neutron or X-ray diffraction Intensityscatteredin the direction k f Equal to the computed structur efactor. Kob et al. 1999

23 Structure factor in multicomponent systems SiO 2, GeSe 2, SiO 2 -Na 2 O, Consider a system with n components havingn 1, N 2, N n particles. We can write Faber-Ziman partial structure factors: = $& (>)= (1+ $&) 2 % % exp [ E. ] out of which can becomputed a total structure factor: Neutron weighted: = H =,( J J ( K K ( = ( (H),( J J ( K K ( withb i the neutron scatteringcross section, c i the concentration of the species b=5.68 i fmfor Te (tabulated, depends on the isotope) X-ray weighted: = H =,( J J ( L (H)L ( (H)= ( (H),( J J ( L (H)L ( ( H) Withf i (k) the X-ray form factor (elastic or inelastic XRD)

24 v-sio 2 l-gese 2 Micoulaut et al., PRB 2009

25 Various levels of agreement between theory and experiments can be found 5 4 Ag25 3 Ag15 S (Q) 2 1 Ag Q (Å -1 ) Micoulaut et al Data Petri, Salmon, 1991

26 Detailed structural analysis from MD Neighbor distribution Remember = 4 First minimum of g(r) can beusedto define the coordination number. But this is an average. Details are provided from the statistical analysis of each atom. Allowsto characterize the nature of the neighborhood Can beextended to partial CN ( = ( 4

27 Examples-1 Amorphous Ge 1 Sb 2 Te 4 Raty et al. Solid State Sciences 2010 Straightforward 4 neighbours around Ge,Si Information on local geometry - short and long bond distance around Ge - CN Te >2

28 Examples-2: Statistics of neighbors with homopolar bonds Ga 11 Ge 11 Te 78 Voleska et al., PRB 2013

29 Detailed structural analysis from MD Bond angle distributions 1173 K SnSe 2 Depending on the system, provides information about the local geometry (tetrahedral, octahedral, ). Directional bonding vs non-directional 300 K SiO 2 -Na 2 O Micoulaut et al. PRB 2008 Cormack and Du, JNCS 2001

30 Detailed structural analysis from MD Ring statistics: serve to characterize the intermediate range order Remember the early work of Galeener (lecture 3) and the Raman characterization of rings Simulated positions can serve to define nodes and links. When connected sequentially without overlap, one has a path. A ring is therefore simply a closed path. Each of these rings is characterized by its size and can be classified. S. Le Roux, P. Jund, Comp. Mater. Sci. 2010

31 Conclusions MD trajectories, once properlygenerated canlead to variousstructural informations under various thermodynamic conditions pair distribution functions g(r) bond distances partial coordination numbers, coordination numbers neighborhood Structure factor S(k) information at intermediate lengthscales Many other quantities of interest Rings Bond angle distributions Topological constraints- > Rigidity transitions (next lectures) With imagination, one can find out much more or invent much more Next lecture : time correlations and linear response

32 Final recommendation regarding structural properties 1. Do the best you can. 2. Don t fool the reader. 3. Be honestwith the data 4. Do not attempt to dissimulate 5. Unique S(k)? Uncarefully simulated liquid GeTe 4 Atomic modeling of glass LECTURE7 MD CALCULATING