Minimize命令Minimize命令使用格式:minimize(命令名)tolerance(误差限)maxiter()maxeval()说明:tolerance=stoppingtolerance停止公差、容差maxiter=优化器的最大叠代次数(maxiterationsofminimizer)maxevel=总的(力/能)评估值的最大个数(maxnumberoftotalforce/energyevalutions)例子:minimze1.0e-41001000描述:该命令用来执行一个系统的能量的最小化(energyminimization),通过调整每个原子的原子坐标。使用的优化算法由命令min_style设置。能量最小化命令可以被和run命令交替使用(interspersed),在松弛和动力学之间分布(Minimizecommandscanbeinterspersedwithruncommandstoalternatebetweenrelaxationanddynamics.)。最小化器areimplementedinarobustfashion坚固的方式thatshouldallowforsystemswithhighlyoverlappedatoms(largeenergiesandforces)tostillbeminimizedbypushingtheatomsoffofeachother.(通过推开每个原子,使得即使有重叠的原子(具有高的能量和力)仍然可以优化,因为算法是很强壮的。不知道这样翻译对不对,这句太难理解了。最优化器involves包括一个外部迭代循环outeriterationloop(用以设置搜索迭代方向,在这个方向上坐标被改变),内部的迭代使用一个线搜索算法(Aninneriterationisthenperformedusingalinesearchalgorithm.T)。线搜寻专门评估力和能量(几次),以设置新的坐标。最小化器在满足下列的任何一个准则的时候停止:*thechangeinenergybetweenouteriterationsislessthanthetolerance*外部迭代的能量的改变小于容许限的时候*thenumberofouteriterationsexceedsmaxiter*外部迭代超出最大迭代次数*thenumberofforceevaluationsexceedsmaxeval*力评估次数超限*the3Ndimensionalforcevectorgoes(nearly)tozero*三维力矢量接近与零对第一个准则,指定的误差是无单位的。当能量delta(即能量之差吧)对能量的幅度magnitude的比等于容差限的时候满足。(e.g.在上面的例子中为onepartin1.0e4,某分之一)在优化期间,外层的迭代次数被当做一个时间步。输出是被这个时间步触发的。即动力学输出或者dump和restart文件等。Outputistriggeredbythistimestep,e.g.thermodynamicoutputordumpandrestartfiles.Foroptimalconvergence对优化收敛性,pair_style最好在能量和力的截断距离上是光滑到零的,尽管不是必须的,但是是最好使用。例子包括了pair/lj/charmm/coul/charmmandpair/lj/charmm/coul/long.。如果一个软势函数被使用,则Astop值用来做预先因子(不和时间有关)IfasoftpotentialisusedtheAstopvalueisusedfortheprefactor(notimedependence).有点晕户,希望明白的后面之解释一下。在优化期间,只有施加了力约束的fix(操作)才被调用。目前的执行表列出:Thelistofthecurrentlyimplementedonesincludefixaddforce,aveforce,enforce2d,indent,lineforce,planeforce,setforce,andwall/lj93。注意indent和wall/lj93有一个联合势能。如果你希望这个能量被包括在系统的总势能中,(属性量正在优化的系统),你必须使得fix_modify命令的能量energy选项对这个fix有效。youmustenablethe[url=file:///D:/My%20Documents/桌面/lammps-11Sep06/doc/fix_modify.html]fix_modify[/url]energyoptionforthatfix.接着优化器的是一个统计的特征总结被打印出来,其中有能量的变化,收敛准则信息等。限制:这里列出来的两个特征还不能做。万一,以防,incase,有些人知道怎么写代码。两个不能被最优化器调用的fix分别是fix_shake和fix_rigid。Itisanerrortousefixshakewithminimizationbecauseitturnsoff(关闭)bondsthatshouldbeincludedinthepotentialenergyofthesystem(应当被包括在系统势能汇总的键).Theeffectofafixshakecanbeapproximated(近似)duringaminimizationbyusingstiffspringconstantsforthebondsand/oranglesthatwouldnormallybeconstrainedbytheSHAKEalgorithm.Fixrigidisalsonotsupportedbyminimization.Itisnotanerrortohaveitdefined(?????),buttheenergyminimizationwillnotkeepthedefinedbody(s)rigid(不能保持定义的刚体???)duringtheminimization.Notethatifbonds,angles,etcinternaltoarigidbodyhavebeenturnedoff(e.g.vianeigh_modifyexclude),theywillnotcontributetothepotentialenergywhichisprobablynotwhatisdesired(可能就不符合实际情况的意思吧).Thevolumeofthesimulationdomainisnotallowedtochange(不许改变)duringaminimzation.Ideally理想的wewouldallowafixsuchasnpttoimposeanexternalpressurethatwouldbeincludedintheminimization(i.e.allowtheboxdimensionstochange),butthishasnotyetbeenimplemented.88888888888888888888888888888888888888下面的是老版本文字的说明Theeffectofafixshakecanbeapproximated近似duringaminimizationbyusingstiffspringconstants刚性弹簧常数forthebondsand/oranglesthatwouldnormally通常,正常beconstrainedbytheSHAKEalgorithm.在最小化期间,模拟域的体积是不容许改变的。理想的,我们容许诸如npt的操作fix来施加一个外部的压力(将被包含在优化器中),即容许模拟盒子的纬度改变,但是这个部分还没有实现。()Thevolumeofthesimulationdomainisnotallowedtochangeduringaminimzation.Ideallywewouldallowafixsuchasnpttoimposeanexternalpressurethatwouldbeincludedintheminimization(i.e.allowtheboxdimensionstochange),butthishasnotyetbeenimplemented相关命令min_modifymin_stylerun_style默认:无min_stylecommand最小化类型命令格式:min_stylestyle*style=cgorcg/frorsd例子min_stylecgmin_stylesd描述:选择一个最小化算法,为minimize命令的执行。类型cg是共轭梯度算法(CG)的Polak_Ribiere(PR)版本)Ateachiteration在每个叠代中,theforcegradientiscombinedwiththepreviousiterationinformation联合前一个叠代的信息tocomputeanewsearchdirection以计算一个新的搜索方向,perpendicular(conjugate)toprevioussearchdirections.和前一个方向是正交的(共轭)。ThePRvariant变化(变量)affectshowthedirectionischosen方向如何选择andhowtheCGmethodisrestartedCG方法如何被重启,whenitceasestomakeprogress当为了进一步而停止的时候.ThePRvariantisthoughttobethemosteffectiveCGchoice.Stylecg/fristheFletcher-Reevesversionoftheconjugategradientalgorithm.共轭梯度算法的另一个版本Stylesdisasteepestdescentalgorithm最速降线法,.Ateachiteration,thedownhilldirection下降方向correspondingtotheforcevector(negativegradientofenergy)issearchedalongbya1dlinesearch沿一个一维线查找.Typically,steepestdescentwillnotconverge不收敛asquicklyasCG,butmaybemorerobustinsomesituations.限制无相关命令min_modifyminimize默认:min_stylecgmin_modifycommand格式:min_modifykeywordvalues...*oneormorekeyword/valuepairsmaybelisted(变量)·keyword=linestyleordminordmaxorlineiter·linestylevalue=secant正割orscan·dminvalue=min·min=minimumdistanceforlinesearchtomove(distanceunits)·dmaxvalue=max·max=maximumdistanceforlinesearchtomove(distanceunits)·lineitervalue=N·N=maxnumberofiterationsinalinesearchExamples:min_modifylinestlescandmin0.001dmax0.2min_modifylineiter5Description:Thiscommandsetsparameters设置参数thataffecttheminimizationalgorithms.Thevarioussettings不同的设置mayeffectth