CHARMM Element doc/ewald.doc 1.1#File: Ewald, Node: Top, Up: (chmdoc/commands.doc), Next: Syntax, Previous: Top The Ewald Summation method Invoking the Ewald summation for calculating the electrostatic interactions can be specified any time the nbond specification parser is invoked. See the syntax section for a list of all commands that invoke this parser. Prerequisite reading: nbonds.doc* Menu:* Syntax:: Syntax of the Ewald summation specification* Defaults:: Defaults used in the specification* Function:: Description of the options* Discussion:: More general discussion of the algorithm#File: Ewald, Node: Syntax, Up: Top, Next: Defaults, Previous: Top[SYNTAX EWALD]{ NBONds } { nonbond-spec }{ UPDAte } { }{ ENERgy } { }{ MINImize } { }{ DYNAmics } { }The keywords are:nonbond-spec::= [ method-spec ] { [ NOEWald ] } { }method-spec::= { EWALd [ewald-spec] { [ NOPMewald [std-ew-spec] ] } } { { PMEWald [pmesh-spec] } } ewald-spec::= KAPPa real [erfc-spec]std-ew-spec::= { [ KMAX integer ] } KSQMAX integer { KMXX integer KMXY integer KMXZ integer }pmesh-spec::= FFTX int FFTY int FFTZ int ORDEr integer [QCOR real (***) ]erfc-spec::= { SPLIne { [EWMIn real] [EWMAx real] [EWNPts int] } } { INTErpolate { } } { } { ABROmowitz } { CHEBychev } { EXACt_high_precision } { LOWPrecision_exact } { ERFMode int }#File: Ewald, Node: Defaults, Up: Top, Next: Function, Previous: SyntaxThe defaults for the ewald summation are set internallyand are currently set to NOEWald, KAPPa=1.0, KMAX=5, KSQMax=27, andNOPMewald, KAPPa=1.0, FFTX=FFTY=FFTZ=32, ORDEr=4, QCOR=1.0Recommended values for Ewald are: EWALD PMEWald KAPPa 0.34 ORDEr 6 - FFTX intboxvx FFTY intboxvy FFTZ intboxvz - CTOFNB 12.0 CUTNB 14.0 QCOR 1.0(***)Where intboxv* is an integer value similar to or larger than the correspondingunit cell dimension that has prime factors of 2,3, and 5 only (2,3 preferred).grid point spacing should be between 0.8 and 1.2 Angstroms.These recommended values should give relative force errors of roughly 10**-5.To reduce the total PME cost at the expense of accuracy, decrease the cutoffdistances while increasing KAPPa (keep the product near 4) reduces the realspace cost. To reduce the K-space cost, either reduce ORDEr from 6 to 4 orincrease the grid spacing up to perhaps 1.5 Angstroms.(***) The QCOR value should be 1.0 for vacuum, solid, or finite systems.For periodic systems in solution, it should be reduced (or set to zero) by anamount that depends on how the net charge is distributed and on the effectivedielectric constant. For a treatise on this correction term, see:S. Bogusz, T. Cheatham, and B. Brooks, JCP (1998) 108, 7070-7084 and referencescontained therein (esp. Hummer and Levy).#File: Nbonds, Node: Function, Up: Top, Previous: Defaults, Next: Discussioni) The EWALD keyword invokes the Ewald summation for calculation of electrostatic interactions in periodic, neutral systems. The formulation of the Ewald summation dictates that the primary system must be neutral. If otherwise, the summation is not formally correct and some convergence problems may result. The NOEWald (default) suppresses the Ewald method for calculating electrostatic interactions. Van der waals options VSHIFT and VSWITCH are supported with ewald. The algorithmcurrently supports the atom and group nonbond lists and the CRYSTAL facilty must be used. The PMEWald keyword invokes the Particle Mesh Ewald algorithmfor the reciprocal space summation. For details on the PME method, seeJ. Chem. Phys. 103:8577 (1995). The EWALd algorithm is limited to CUBIC,TETRAGONAL, and ORTHORHOMBIC unit cells. The PMEWald algorithm supportsall unit cells that are supported by the CRYSTAL facility. ii) The KAPPa keyword, followed by a real number governs the width of the Gaussian distribution central to the Ewald method. An approximate valueof kappa can be chosen by taking KAPPa=5/CTOFNB. This is fairly conservative.Values of 4/CTOFNB lead to small force errors (roughly 10**-5). Seediscussion section for details on choosing an optimum value of KAPPa.iii) The KMAX key word is the number of kvectors (or images of the primary unit cell) that will be summed in any direction. It is the radius of the Ewald summation. For orthorombic cells, the value of kmax may be independently specified in the x, y, and z directions withthe keywords KMXX, KMXY, and KMXZ. In the PME version, the number of FFT grid points for the charge mesh is specified by FFTX, FFTY, and FFTZ. iv) The KSQMax key word should be chosen between KMAX squared and 3 timesKMAX squared. v) An appropriate, although not optimal, set of parameters can be chosen by taking KAPPA=5/CTOFNB and KMAX=KAPPa*boxlength. The actual values should then be performanced optimized for your particular system.For the PME method, FFTX should be approximately the box length in Angstroms.(for efficiency, FFTX should be a multiple of powers of 2,3, and 5).IMPORTANT NOTE::: THE SUGGESTION THAT FFTX, FFTY, AND FFTZ HAVE NO PRIME FACTORS OTHER THAN 2, 3, AND 5 SEEMS TO BE A REQUIREMENT. LARGE ERRORS IN THE FORCE ARE OBSERVED WHEN THIS CONDITION IS NOT MET. FUTURE VERSIONS OF CHARMM WILL FLAG THIS AS AN ERROR CONDITION.ORDEr specifies the order of the B-spline interpolation, e.g. cubic isorder 4 (default), fifth degree is ORDEr 6. The ORDEr must be an evennumber and at least 4. vi) EWALd runs in parallel on both shared (PARVECT) and distributedmemory parallel computers. PME runs in parallel on distributedmemory computers.vii) several algorithms are available for the calculation of the complimentaryerror function, erfc(x).