• Keine Ergebnisse gefunden

Coulomb interactions: P3M, MMMxD, ELC and ICC

N/A
N/A
Protected

Academic year: 2021

Aktie "Coulomb interactions: P3M, MMMxD, ELC and ICC"

Copied!
27
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)Coulomb interactions: P M, MMMxD, ELC and ICC∗. http://www.icp.uni-stuttgart.de. 3. Axel Arnold Institute for Computational Physics Universität Stuttgart October 10, 2012.

(2) Electrostatics Coulomb energy Pair energy summation. Bjerrum length. http://www.icp.uni-stuttgart.de. N. l X0 qi qj U= B 2 |rij |. lB =. e2 4π0 r kB T. i,j=1. summing up 1/r Coulomb pair potential. electrostatic prefactor ∝ inverse temperature. Bjerrum length lB. for two unit charges:. 1kBT 1lB. A. Arnold. Coulomb interactions. 2/26.

(3) Electrostatics Coulomb energy Pair energy summation. http://www.icp.uni-stuttgart.de. U=. N lB X0 qi qj 2 |rij |. Potential summation N. U=. i=1. i,j=1. summing up 1/r Coulomb pair potential. 1X qi φ(ri ) 2. potential from solving Poisson’s equation. Bjerrum length lB ∇2 φ(r) = −4πlB. N X. qj δ(rj − r). j=1. equivalent approaches A. Arnold. Coulomb interactions. 2/26.

(4) Electrostatics in periodic boundary conditions Coulomb energy. http://www.icp.uni-stuttgart.de. Pair energy summation. Potential summation. N ∞ qi qj lB X X X 0 U= 2 |r ij + mL| 2. N. U=. i=1. S=0 m =S i,j=1. conditionally convergent — summation order important. 1X qi φper (ri ) 2. solve Poisson’s equation imposing periodic boundaries. numerically difficult U not periodic in coordinates ri. U is periodic in coordinates ri. these two calculate something different!. A. Arnold. Coulomb interactions. 3/26.

(5) Where the difference comes from: the dipole term assume summation in periodic shells surrounded by polarizable material of dielectric constant ∞  = ∞. http://www.icp.uni-stuttgart.de. 3. 3 2. 1. 1. 0. 1. 2. 1. 2. =1. 3. Pair energy summation vacuum around: ∞ = 1. 2 3.  = ∞. Potential summation periodic: ∞ = ∞. difference to periodic solution is nonperiodic dipole term X 2 2π (d) U = qi ri (1 + 2∞ )L3 i. metallic boundary conditions ∞ = ∞ always safe never use ∞ < ∞ for conducting systems A. Arnold. Coulomb interactions. 4/26.

(6) Electrostatics in ESPResSo requires myconfig.h-switch ELECTROSTATICS switching on: inter coulomb <lB > < method > < parameters >. methods and their parameters: next 2 hours http://www.icp.uni-stuttgart.de. switching off: inter coulomb 0. getting lB , method and parameters: inter coulomb. returns e. g. { coulomb 1 .0 p3m 7 .75 8 5 0 .1138 0 .0 } { coulomb epsilon 80 .0 n_interpol 32768 mesh_off 0 .5 0 .5 0 .5 }. A. Arnold. Coulomb interactions. 5/26.

(7) Assigning charges part 0 pos 0 0 0 q 1 part 1 pos 0 .5 0 0 q -1.5. Adding a charged plate. http://www.icp.uni-stuttgart.de. constraint plate height <h> sigma <σ >. plate parallel to xy-plane at z = h, charge density σ requires 2D periodicity (nonperiodic in z). Adding a charged rod constraint rod center <cx > <cy > lambda <λ>. rod parallel to z-axis at (x, y) = (cx , cy ), line charge density λ requires 1D periodicity (periodic in z) A. Arnold. Coulomb interactions. 6/26.

(8) P. P. Ewald, 1888 — 1985 Coulomb potential has 2 problems 1. singular at each particle position 2. very slowly decaying Idea: separate the two problems! one smooth potential — Fourier space one short-ranged potential — real space A. Arnold. Coulomb interactions. P. P. Ewald, Die Berechnung optischer und elektrostatischer Gitterpotentiale, Ann. Phys. 369(3):253, 1921. http://www.icp.uni-stuttgart.de. The Ewald method. 7/26.

(9) Ewald: splitting the potential charge distribution ρ=. N X X. qi δ(r − ri − n). http://www.icp.uni-stuttgart.de. n∈LZ3 i=1. =. +. replace δ by Gaussians of width α−1 : √ 3 2 2 ρGauss (r) = α/ π e−α r δ(r) = ρGauss (r) + [δ(r) − ρGauss (r)] A. Arnold. Coulomb interactions. 8/26.

(10) The Ewald formula U = U (r ) + U (k) + U (s) with U (r ) =. erfc(α|rij + mL|) lB X X0 qi qj 2 |rij + mL| 3 i,j. m∈Z. http://www.icp.uni-stuttgart.de. real space correction. U (k). l X 4π −k 2 /4α2 = B3 e |b ρ(k)|2 2L k2. U (s). αl X 2 = − √B qi π. Gaussians in k -space. k6=0. Gaussian self interaction. i. forces from differentiation ∂ U ∂ri ... coming soon to ESPResSo (on GPU) Fi = −. A. Arnold. Coulomb interactions. 9/26.

(11) discrete FT is exact — constant real space cutoff computational order O(N log N) most frequently used methods: P3 M: optimal method PME SPME A. Arnold. Coulomb interactions. M. Deserno and C. Holm, JCP 109:7678, 1998. R. W. Hockney J. W. Eastwood replace k -space Fourier sum by discrete FFT. R. W. Hockney and J. W. Eastwood, Computer Simulation Using Particles, 1988. http://www.icp.uni-stuttgart.de. Mesh-based Ewald methods. 10/26.

(12) Steps of P3 M 1. {ri , qi } → ρ(r): interpolate charges onto a grid (window functions: cardinal B-splines). http://www.icp.uni-stuttgart.de. 2. ρ(r) → ρb(k):. Fourier transform charge distribution. 3. φ̂(k) = Ĝ(k)ρ̂(k): solve Poisson’s equation by multiplication with optimal influence function Ĝ(k) (in continuum: product of Green’s function 4π and k2 2 /4α2 −k Fourier transform of Gaussians e ) 4. ikφ̂(k) → Ê(k):. obtain field by Fourier space differentiation. b 4. E(k) → E(r):. Fourier transform field back. 5. E(r) → {ri , Fi }: interpolate field at position of charges to obtain forces Fi = qi Ei A. Arnold. Coulomb interactions. 11/26.

(13) Charge assignment q7. q8. 1. 1. 2. 0.8. q5. 3 4. M (P ) (x). q6. 5. 0.6. 6. 7. 0.4. q3. http://www.icp.uni-stuttgart.de. q1. a. q4. 0.2. q2. 0. 0. 1. 2. 3. 4. 5. 6. 7. x. interpolate charges onto h-spaced grid ρM (rp ) =. N 1 X qi W (p) (rp − ri ) h3 i=1. W (p) (r) cardinal B-splines in P3 M / SPME A. Arnold. Coulomb interactions. 12/26.

(14) Optimal influence function. Ĝopt (k) = h. 6 ik. ·. P. 2.  |k|2. http://www.icp.uni-stuttgart.de. 2π 2π b h m)R(k + h m)  2 2 P [ (p) (k + 2π m) W m∈Z3 h. m∈Z3. [ (p) (k + W. aliasing of continuum force 4π 2 2 b R(k) = −ik 2 e−k /4α k with differentiation, Green’s function and transform of Gaussian minimizes the rms force error functional Z Z  2 1 3 Q[F ] := 3 d r1 d3 r F(r; r1 ) − R(r) h h3 V A. Arnold. Coulomb interactions. 13/26.

(15) Why to control errors rms force error ∆F =. q. s (Fexact − FEwald )2 =. 10. 1 N. N P i=1. ∆Fi2. rmax=1, kmax=10 rmax=2, kmax=10 rmax=1, kmax=20. 1. http://www.icp.uni-stuttgart.de. ∆F. 0.1 0.01 0.001 0.0001 1e-05 0. 1. 2. 3. 4. 5. α. optimal α brings orders of magnitude of accuracy at given required accuracy, find fastest cutoffs compare algorithms at the same accuracy. A. Arnold. Coulomb interactions. 14/26.

(16) How to: error estimates 10. total error real space estimate k-space estimate. 1. ∆F. 0.1 0.01 0.001 0.0001. http://www.icp.uni-stuttgart.de. 0. 1. 2. 3. 4. 5. α. Kolafa and Perram: ∆Freal. P 2   q 2 2 ≈ √ i p exp −α2 rmax N rmax L3. Hockney and Eastwood: ∆FFourier A. Arnold. P 2 q ≈ √ i N. s Q[Ĝopt (k)] L3. Coulomb interactions. 15/26.

(17) P3 M in ESPResSo (F. Weik, H. Limbach, AA) tune P3 M for rms force error τ. http://www.icp.uni-stuttgart.de. inter coulomb <lB > p3m tune accuracy <τ > \ [ r_cut <rmax >] [ mesh <nM >] [ cao <p >] inter coulomb epsilon ∞. computes optimal α tunes for optimal speed rmax real space cutoff (0 to retune) nM = L/h mesh size (0 to retune) p charge assignment spline order p (0 to retune) fixing parameters speeds up tuning, if you know the optimal value second command to set ∞ (defaults to ∞ (“metallic”)) manually set parameters (dangerous!) inter coulomb <lB > p3m <rmax > <nM > <p> <α>. A. Arnold. Coulomb interactions. 16/26.

(18) http://www.icp.uni-stuttgart.de. Partially periodic systems. partially p. b. c. for slablike systems (surfaces, thin films) ... or for cylindrical systems (rods, nanopores) dielectric contrasts at interfaces P3 M cannot be employed straightforwardly. A. Arnold. Coulomb interactions. 17/26.

(19) Another approach: MMM2D far formula φβ (r ) =. e−β. X. √. (x+k )2 +(y +l)2 +z 2. p (x + k)2 + (y + l)2 + z 2 k ,l∈LZ    q q 2 X X 2 2 2 2 = K0 β +p (y + l) + z  eipx L 2π p∈ L Z. http://www.icp.uni-stuttgart.de. =. 2π L2. l∈LZ. X p,q∈ 2π Z L. √ 2 2 2 e− β +p +q |z| ipx iqy p e e 2 2 β + p + q2.  2π  X = 2 L 2 2. p +q >0.  π efpq |z| ipx iqy e e + |z| + 2 β −1 + Oβ→0 (β) fpq L. screened Coulomb interaction in limit of screening length ∞ other formula for z ≈ 0 optimal computation time O(N 5/3 ), comparable to Ewald analogously for 1d, but then O(N 2 ) A. Arnold. Coulomb interactions. 18/26.

(20) Dielectric contrasts z. water. electrode. q∆t ∆b εt q ∆t. membrane. lz. q. http://www.icp.uni-stuttgart.de. q ∆b. εm. x. εb. q ∆b∆t. water. water. electrode. typical two dimensional systems: thin films, slit pores material boundaries =⇒ dielectric contrast take into account polarization by image charges can be handled by MMM2D A. Arnold. Coulomb interactions. 19/26.

(21) MMM2D and MMM1D in ESPResSo (AA) using MMM2D tuned for maximal pairwise error τ. http://www.icp.uni-stuttgart.de. cellsystem layered <nlayers > inter coulomb <lB > mmm2d <τ > [ <kmax >] \ [ dielectric <t > <m > <b > | d i e l e c t r i c - c o n t r a s t s <∆t > <∆b >]. allows to fix kmax (p, q)-space cutoff t , m , b dielectric constants or ∆t , ∆b dielectric contrasts requires layered cell system number of layers per CPU nlayers = B/Np is tuning parameter using MMM1D tuned for maximal pairwise error τ cellsystem nsquare inter coulomb <lB > mmm1d tune <τ >. requires all-with-all cell system A. Arnold. Coulomb interactions. 20/26.

(22) The method of Yeh+Berkowitz replicated slab system slab system h Lz. http://www.icp.uni-stuttgart.de. replicated slab system L. potential of a charge and its periodic images similar to plate plates cancel due to charge neutrality 2πqi. N X. σj (|zji + mLz | + |zji − mLz |) = 4πqi nLz. j=1. N X. σj = 0. j=1. leave a gap and hope artificial replicas cancel  2 2π P (d) requires changed dipole term U = L3 i qi zi A. Arnold. Coulomb interactions. 21/26.

(23) Ulc =. N e|k|zij + e−|k|zij i(kx xij + ky yij ) π X X qi qj e 2 L fpq (efpq Lz − 1) 2π 2 k∈ L Z k2 >0. i,j=1. error not known a priori — required gap size? calculate contribution of image layers subtract numerically ⇒ needs smaller gaps 2-4x faster than plain Yeh+Berkowitz A. Arnold. Coulomb interactions. AA, J. de Joannis, and C. Holm, JCP 117:2496, 2002. http://www.icp.uni-stuttgart.de. Electrostatic layer correction (ELC). 22/26.

(24) ELC in ESPResSo (AA) using ELC for maximal pairwise error τ inter coulomb <lB > p3m tune accuracy <τ 0 > ... inter coulomb elc <τ > <g > [kmax >] \ [ dielectric <t > <m > <b > | d i e l e c t r i c - c o n t r a s t s <∆t > <∆b >]. http://www.icp.uni-stuttgart.de. gap size g = Lz − h has to be specified user is responsible to keep a gap (by walls or fixed particles) gap location unimportant requires P3 M to be switched on first allows to fix kmax t , m , b ∆t , ∆b. A. Arnold. (p, q)-space cutoff (otherwise tuned) dielectric constants or dielectric contrasts. Coulomb interactions. 23/26.

(25) http://www.icp.uni-stuttgart.de. Arbitrarily shaped dielectric surfaces. MMM2D/ELC only handle planar parallel dielectric interfaces what about a nanopore? vesicle? cannot be handled by image charges satisfy boundary constraints for electric field. A. Arnold. Coulomb interactions. 24/26.

(26) ICC∗ algorithm boundary condition at interface. http://www.icp.uni-stuttgart.de. εin Ein · n = εout Eout · n. nk. Ak. can be fulfilled by interface charge density σ=. ε − εout 1 εout in E(σ) 2π εin + εout. represent σ as charges qk at fixed positions at interface solve for qk iteratively, E from standard Coulomb solver   ε − εout qkl+1 = (1 − ω) qkl + ωAk in nk · E [qjl ] εin + εout A. Arnold. Coulomb interactions. 25/26.

(27) ICC∗ in ESPResSo (S. Kesselheim) set up the meshed interfaces. http://www.icp.uni-stuttgart.de. dielectric sphere center <x y z > \ radius <r > res <a> eps <εin >. a is mesh size of the generated mesh alternatively wall, pore, cylinder creates Tcl variables with properties of the surface points: n induced charges icc epsilons: list of dielectric constants, can vary per surface point icc normals: list of normal vectors icc areas: list of surface areas sigmas: optional list of additional surface charge densities. surfaces charges are calculated by iccp3m $ n _ i n d u c e d _ c h a r g e s epsilons $icc_epsilons \ normals $icc_normals areas $icc_areas [ sigmas $icc_sigmas ]. A. Arnold. Coulomb interactions. 26/26.

(28)

Referenzen

ÄHNLICHE DOKUMENTE

The quality of the batches of drug substance placed into the stability program should be representative of the quality of the material used in preclinical and clinical studies and

Das Ziel dabei heißt, Erster zu sein mit einer Idee, diese als Erster zu kom- munizieren und sich damit als erster in den Köpfen der Zielgruppe festzusetzen.. Dieses Ziel

Heute sind es sieben Patienten, die mit ihrem Trinkverhalten nicht nur ihre Gesundheit, sondern auch ihren Beruf und ihre Familie aufs Spiel setzen.. Und diese

Dann habe ich mich aber doch für einen Burn-on entschieden: Auf der Flamme des feu sacré möchte ich genüsslich verbrennen, zum Wohl meiner Patient- Innen, meiner Lieben und

[r]

[r]

Suche dir für q selbst eine Farbe zum

at the same time, the discourse has given relatively little discussion to some proven technologies, particularly dry cask storage, that can provide South Korea ways to manage