am C CASP 6.0 AlstonJ.Misquitta†andAnthonyJ.Stone†† †DepartmentofPhysicsandAstronomy, andtheThomasYoungCentreforTheoryandSimulationofMaterials, QueenMaryUniversityofLondon,327MileEndRoad,LondonE14NS ††UniversityChemicalLaboratory,LensfieldRoad,CambridgeCB21EW February28,2020 Abstract CamCASP is a suite of programs designed to calculate molecular properties (multipoles and frequency- dependentpolarizabilities)insingle-siteanddistributedform,andinteractionenergiesbetweenpairsofmolecules, and thence to construct atom–atom potentials. The CamCASP distribution also includes the programs PFIT, Casimir,GDMA2.3,Cluster,andProcess. Copyright(cid:13)c 2007–2018AlstonJ.MisquittaandAnthonyJ.Stone Contents 1 Introduction 1 1.1 Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Citations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 What’snew? 2 3 OutlineofthecapabilitiesofCamCASP andotherprograms 3 3.1 CamCASP limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 4 Installation 5 4.1 BuildingCamCASP fromsource . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 5 UsingCamCASP 9 5.1 Workflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 5.2 High-levelscripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 5.3 Theruncamcasp.pyandsubmit_camcasp.pyscripts . . . . . . . . . . . . . . . . . . . . . . . . 10 5.4 Low-levelscripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 6 Dataconventions 12 7 CLUSTER:Detailedspecification 13 7.1 Prologue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 7.2 Moleculedefinitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 7.3 Geometrymanipulationsandothertransformations . . . . . . . . . . . . . . . . . . . . . . . . 15 7.4 Jobspecification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 7.5 Energy&Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 7.6 Crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 7.7 ORIENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 7.8 Finally,... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 8 Examples 35 8.1 ASAPT(DFT)calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 8.2 Anexamplepropertiescalculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 8.3 Dispersioncoefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 8.4 UsingCLUSTERtoobtainthedimergeometry . . . . . . . . . . . . . . . . . . . . . . . . . . 43 9 CamCASPprogramspecification 45 9.1 Globaldata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 9.2 Moleculedefinition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 9.3 TheEDITmodule:Modifyingamolecularspecification . . . . . . . . . . . . . . . . . . . . . . 48 9.4 Density-fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 9.5 Propagatorsettings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 9.6 Quadraturesettings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 9.7 TheISAmodule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 9.8 Multipolemoments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 9.9 Display . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 9.10 Polarizabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 9.11 Lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 9.12 Numericalintegrationgrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 9.13 Electrostaticinteraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 9.14 First-orderexchange(exchange-repulsion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 9.15 Second-orderinductionenergy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 9.16 Second-orderdispersionenergy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 9.17 Distributeddensity-overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 9.18 Energyscan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 9.19 Overlapmodel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 9.20 Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 10 PROCESS:Syntax 82 11 CASIMIR 86 A Basissets 88 B Dispersioncoefficients:indetail 89 C Changehistory 91 D OlderScripts 94 D.1 High-levelscripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 D.2 Low-levelscripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 3 1 Introduction CamCASP is a suite of programs for the calculation of interaction energies between pairs of molecules, and molecularproperties(multipolesandfrequency-dependentpolarizabilities)insingle-siteanddistributedform.The CamCASP distributionalsoincludestheprogramsPFIT,Casimir,GDMA2.3,Cluster,andProcess,andtogether theseformapackagefortheabinitiogenerationofsite–siteforcefieldsbetweenorganicmoleculescontainingup toabout60atoms. 1.1 Authors The CamCASP suite of programs, which includes PFIT, Casimir, GDMA 2.3, Process, and Cluster, has been writtenbyAlstonJ.MisquittaandAnthonyJ.StonewithimportantcontributionsfromRobertBukowski,Wojciech Cencek,theGAMESS(US)teamandtheGAUSSINTteam. 1.2 Citations Thiscodeisprovidedasaservicetothescientificcommunityandouronlyrecompense,suchasitis,isincitations. Therefore,ifyouuseanyresultsfromCamCASP inyourpublicationswerequestthatyoucitethefollowingpapers. Thechoiceofcitationswoulddependonthepartsofthecodeyouhaveusedforthepublishedresults. • SAPT(DFT)energies ◦ A.J.MisquittaandK.Szalewicz. Intermolecularforcesfromasymptoticallycorrecteddensityfunc- tionaldescriptionofmonomers. Chem.Phys.Lett.,357:301–306,2002 ◦ A. J. Misquitta, B. Jeziorski, and K. Szalewicz. Dispersion energy from density-functional theory descriptionofmonomers. Phys.Rev.Lett.,91:33201,2003 ◦ A.J.MisquittaandK.Szalewicz. Symmetry-adaptedperturbation-theorycalculationsofintermolecu- larforcesemployingdensity-functionaldescriptionofmonomers. J.Chem.Phys.,122:214109,2005 ◦ A. J. Misquitta, R. Podeszwa, B. Jeziorski, and K. Szalewicz. Intermolecular potentials based on symmetry-adaptedperturbationtheorywithdispersionenergiesfromtime-dependentdensity-functional theory. J.Chem.Phys.,123:214103,2005 ◦ R.Bukowski,R.Podeszwa,andK.Szalewicz.EfficientgenerationofthecoupledKohn–Shamdynamic susceptibility functions and dispersion energy with density fitting. Chem. Phys. Lett., 414:111–116, 2005 • WSMpolarizabilities ◦ A.J.MisquittaandA.J.Stone. Distributedpolarizabilitiesobtainedusingaconstraineddensity-fitting algorithm. J.Chem.Phys.,124:024111,2006 ◦ A.J.MisquittaandA.J.Stone. Accurateinductionenergiesforsmallorganicmolecules:I.Theory. J.Chem.TheoryComput.,4:7–18,2008a ◦ A.J.Misquitta,A.J.Stone,andS.L.Price. Accurateinductionenergiesforsmallorganicmolecules. 2.DevelopmentandtestingofdistributedpolarizabilitymodelsagainstSAPT(DFT)energies.J.Chem. TheoryComput.,4:19–32,2008a. doi:10.1021/ct700105f • WSMDispersionmodels ◦ A. J. Misquitta and A. J. Stone. Dispersion energies for small organic molecules: first row atoms. Molec.Phys.,106:1631–1643,2008b • SRLOpolarizabilities ◦ A.J.MisquittaandA.J.Stone. Distributedpolarizabilitiesobtainedusingaconstraineddensity-fitting algorithm. J.Chem.Phys.,124:024111,2006 ◦ FazleRobandKrzysztofSzalewicz. Asymptoticdispersionenergiesfromdistributedpolarizabilities. Chem.Phys.Lett.,572:146–149,2013 1 • GDMAmultipolemoments ◦ A.J.Stone. Distributedmultipoleanalysis:Stabilityforlargebasissets. J.Chem.TheoryComput.,1: 1128–1132,2005 • Potentials&Overlapmodels ◦ A. J. Stone and A. J. Misquitta. Atom–atom potentials from ab initio calculations. Int. Rev. Phys. Chem.,26:193–222,2007 ◦ A.J.Misquitta,G.W.A.Welch,A.J.Stone,andS.L.Price. Afirstprinciplespredictionofthecrystal structureofC Br ClFH . Chem.Phys.Lett.,456:105–109,2008b 6 2 2 ◦ AlstonJ.MisquittaandAnthonyJ.Stone. Abinitioatom-atompotentialsusingcamcasp:Theoryand applicationtomany-bodymodelsforthepyridinedimer. J.Chem.TheoryComput.,12(9):4184–4208, 2016.doi:10.1021/acs.jctc.5b01241.URLhttps://doi.org/10.1021/acs.jctc.5b01241.PMID: 27467814 • Charge-tranferviaregularisation ◦ A. J. Misquitta. Charge-transfer from regularized symmetry-adapted perturbation theory. J. Chem. TheoryComput.,9:5313–5326,2013. doi:10.1021/ct400704a • Iteratedstockholderatom(ISA) ◦ AlstonJ.Misquitta,AnthonyJ.Stone,andFarhangFazeli. Distributedmultipolesfromarobustbasis- space implementation of the iterated stockholder atoms procedure. J. Chem. Theory Comput., 2014. doi:10.1021/ct5008444 ◦ A. J. Misquitta and A. J. Stone. Isa-pol: Distributed polarizabilities and dispersion models from a basis-spaceimplementationoftheiteratedstockholderatomsprocedure. Theor.Chim.Acta,137:153 (20),2018 • ISA-Poldistributedpolarizabilitiesanddispersionmodels ◦ A. J. Misquitta and A. J. Stone. Isa-pol: Distributed polarizabilities and dispersion models from a basis-spaceimplementationoftheiteratedstockholderatomsprocedure. Theor.Chim.Acta,137:153 (20),2018 Furthermore,ifyoumakeanychangesoradditionstothecodeandwouldliketosharethemforinclusioninfuture releases,pleasesubmitthemodificationstouswithsuitabledocumentationandexamples. 2 What’s new? Thecurrentversionis6.0.xx.Enhancementstothecodesinceversion5.9: • E(2) withouttheS2approximation.Implementationofaclosed-shellversionoftheformulaefromSchäf- ind,exch ferandJansen[SchäfferandJansen,2012]. • NewversionoftheBS-ISAalgorithm.Moreflexibilityinthebasissetsandconsequenthigheraccuracies. DetailsofthisalgorithmaredescribedbyMisquittaandStone[MisquittaandStone,2018]. • ISA-Pol: ISA-based distributed polarization code that computes frequency-dependent distributed polariz- abilitiesanddispersionmodels.SeeA.J.MisquittaandA.J.Stone. Isa-pol:Distributedpolarizabilitiesand dispersion models from a basis-space implementation of the iterated stockholder atoms procedure. Theor. Chim.Acta,137:153(20),2018foradescription. • Distributeddensity-overlapwiththeISA.SeeAlstonJ.MisquittaandAnthonyJ.Stone.Abinitioatom-atom potentials using camcasp: Theory and application to many-body models for the pyridine dimer. J. Chem. Theory Comput., 12(9):4184–4208, 2016. doi: 10.1021/acs.jctc.5b01241. URL https://doi.org/10. 1021/acs.jctc.5b01241. PMID:27467814fordetails. 2 • Advances in the fitting strategy for the development of atom–atom interaction models. See Alston J. Mis- quitta and Anthony J. Stone. Ab initio atom-atom potentials using camcasp: Theory and application to many-bodymodelsforthepyridinedimer. J.Chem.TheoryComput.,12(9):4184–4208,2016. doi:10.1021/ acs.jctc.5b01241. URLhttps://doi.org/10.1021/acs.jctc.5b01241. PMID:27467814fordetails. • Interface with Psi4 1.1 and the unreleased github versions. Both SAPT(DFT) and δHF energies can be int computedusingPsi4 asthefront-endcode. • The introduction of methods for important types of calculations with CamCASP . These methods allow complexcalculationsinaneasywayandalsoallowtheusertodevelopandmaintaintheirownvariationson theprovidedmethods. • IntegrationwithOrient4.9. • ManymorekindsofoperationspossiblewiththeClustercode. • Basissets:ThebasissetlibraryincludedwithCamCASP hasbeensignificantlyexpanded.Whenusedwith Psi4 themainbasissetsaretakenfromthePsi4 basislibrary.WehavealsoincludedtheaugA-Sadlejbasis which is a modified version of the Sadlej-pVTZ basis that results in interaction energies of much higher quality(i.e.,closertotheCBSlimit),particularlywhenusedwiththelargerofthemid-bondsets.Thisbasis ishighlyrecommended. Formoredetails,seetheChangeLogfile. EarlierchangesarelistedinAppendixC. 3 Outline of the capabilities of CamCASP and other programs ThefollowingtypesofcalculationarepossiblewithCamCASP: • Dimerenergies: Thefirst-orderelectrostaticandexchangeenergies,E(1) andE(1) ,andthesecond-ordertotaldispersionand elst exch totalinductionenergies,E(2) andE(2) (includingtheirexchangeterms).Allarecalculatedusingdensity- IND DISP fitting,andthesecond-orderdispersioncaninprinciplebecalculatedwithoutdensity-fitting,thoughthispart ofthecodehasnotbeentestedinalongtime.TheδHFcorrectioncanalsobecalculated. int • Molecularproperties: ◦ Multipole moments: Total and distributed multipole moments can be calculated using a constrained density-fittingalgorithmandtheGDMA2.3code,whichhasbeeninterfacedtoCamCASP. ◦ Frequency-dependent polarizabilities: Total and distributed frequency-dependent polarizabilities are calculatedusingaconstraineddensity-fittingalgorithm.TheWilliams–Stone–Misquitta(WSM)method canbeusedtoobtainthemostaccuratepolarizabilitymodelwithinconstraintsimposedbytheuser. ◦ Point-to-pointpolarizabilities:Responsestoafrequency-dependentpoint-chargeperturbation—called point-to-pointpolarizabilities—canbecalculated.TheseareneededbythePFITprogramwhenopti- mizingthedistributedpolarizabilities. TheOrientprogramisneededforsomeaspectsofthepropertycalculations.Itisdistributedseparatelybut accesstoitisprovidedusingthesamecredentialsasforCamCASP. • Energyscans: Surfaces around a molecule can be defined (or supplied as a grid of points) and all dimer energies can be calculatedonthissurface,usingapointcharge(fortheinductionandelectrostaticsonly)oraprobemolecule. Thetotalanddistributedcharge-densityoverlapisalsoscanned. • Atom–atompotentials: Usingtheresultsoftheenergyscansandmolecularproperties,atom–atompotentialscanbeobtainedusing thedensityoverlapmodeltomodeltheshort-rangeenergies.TheOrientprogramisusedforsomepartsof thiscalculation. 3 • Theorylevels: Inallcases,calculationscanbeperformedusingeitherthecoupled/uncoupledKohn–Sham(CKS/UCKS)or coupled/uncoupledHartree–Fockpropagators.Inpractice,theCKSpropagatorwillbeused. • Density-fittedIntegralPackage: The density-fitted (DF) integral package allows the calculation of 2-electron 4-index integrals in very effi- ciently, both in computation time and memory usage. With this package, the intermediate TRAN step is no longerneededfortheHessiancalculation,therebysavingalotoftime.Inprinciple,theSapt2008program canalsoberunusingintegralsobtainedfromthispackage,butthiswouldinvolvesomemoreeffort. Atpresent,theDF-integralpackagecomputesmorethan30kindsofintegral. ThefunctionalityofCamCASP isgreatlyenhancedwhenusedwiththePFITandOrientprograms.Onlysome ofthemanyfeaturesofOrient—thosemostusefulforthecalculationandanalysisofintermolecularforces—are listedhere. • PFIT ◦ Fitting of polarizability models: At present, polarizability models are not fitted from scratch, but po- larizability models from the CamCASP code can be optimised using PFIT and the point-to-point polarizabilitiesfromCamCASP. ◦ Dispersioncoefficients:ThesecanbecalculatedbyPFITusingfrequency-dependentdistributedpolar- izabilitiesfromCamCASP.ThisfunctionisbetterperformedbytheCasimirprogram. • Casimir ◦ Dispersioncoefficients:C ,C ,etc.toC .Dispersioncoefficientsbetweenpairsofidenticalordissim- 6 7 12 ilarmoleculescanbecalculatedbyCasimirusingfrequency-dependentdistributedpolarizabilitiesfrom CamCASP thathavebeenlocalizedusingOrientandpossiblyrefinedusingPFIT.Theinputfilefor CasimircanbecreatedwiththeProcessprogramfromlocalizedfrequency-dependentpolarizabilities. Minorchangestotheinputfilesallowthecalculationofmixed-moleculedispersioncoefficients. • Dispersion ◦ Dispersionenergiesusingnon-localpolarizabilities:Thisprogramallowstwoandthree-bodydisper- sionenergycalculationsusingfrequency-dependentnon-localpolarizabilitiessuchasthosecalculated using the distributed polarizability module in CamCASP . These calculations include contributions fromthecharge-flowpolarizabilities. • Orient ◦ Local polarizabilities: The distributed polarizabilities from CamCASP include non-local contribu- tions.ThesecanbetransformedawayusingthelocalizationmoduleinOrient. ◦ Simplificationofmodels:ThepolarizabilitymodelscanbemodifiedorsimplifiedusingOrient. ◦ Displayingenergies:Energiescanbedisplayedin3-DusingtheOpenGLdisplaymoduleinOrient. ◦ Calculationsofasymptoticenergies:Usingthedistributedmultipolesandpolarizabilities,interaction energiescanbecomputedinthelong-range(asymptotic)approximation. Miscellaneousprograms: • Process:ThiscodeisaninterfacecodetoprocesspolarizabilitiesfromCamCASP or(localized/modified) polarizabilitiesfromOrientintoaformPFITcanuse.Italsoperformstransformationsofthepolarizabilities andisabletowriteoutpolarizabilitiesinLATEXformatforpublications. • Cluster:Thisistheprogramthathandlestheuser’sinputdataandgeneratesthesubsidiarydatafilesneeded byotherpartsofthepackageandbytheauxiliaryabinitiocodes.Italsoenablestheusertoconductelemen- tarymanipulationsofthemoleculegeometry,buildclusters,determinerotationalandtranslationaxes,and writeoutputinaformsuitableforCamCASP,Orient,andanyprogramthatcanreadPDBfilesforimaging. An important function of Cluster is to create the input files for CamCASP , DALTON (versions 2.0 and 2013orlater),NWChem6.x,Psi4 andSapt2008programs.Alltheuserneeddoisdefinethedimerwithin 4 Cluster,performanymanipulationsthatmightbeneeded,andthenusetheRUN-TYPEcommandtogenerate the files for the calculation. This greatly simplifies the file generation process and removes the chance of errors in the rather complicated input file structure of Sapt2008, DALTON and (to a lesser extent) Psi4 , NWChem6.xandCamCASP. 3.1 CamCASP limits CamCASP isaresearchcodeandwhileitisefficientlycoded,itisnotasyetparallelised,socalculationsonlarge systemscanbechallenging.Neverthelesscalculationsonmoderatelylargedimersarepossible:fortheRDXdimer (30 heavy atoms and 6 hydrogen atoms) using PBE0 with the ALDA kernel and the 6.0 version of CamCASP , theSAPT(DFT)calculationtakes17honasinglecoreandtheDFTcalculationsneededforthistake1h20musing thePsi4 coderunningon4cores.IfthemoreaccuratebutmorecomputationallydemandingALDA+CHFkernel is used, then the SAPT(DFT) calculation takes 22h on 1 core. The differences in memory used are larger: the ALDA calculation used 30GB while the ALDA+CHF calculation used 41GB. These timings and memory sizes werefortheaugA-Sadlejmainbasisandaug-cc-pVTZ-RIdensity-fittingbasiswiththe3s3p2d2fmid-bondsetand correspondingRIbasis.Thisgivesamainbasissizeof803functions,andauxiliarybasissizeof4805functions. The followinglimits applyonly tothose partsof the calculationthat requireintegrals ofthe type OVOV or OOVV. Theseoccurinthehybridkernelsandinthesecond-orderexchange-dispersionandexchange-inductionenergies. Ifonlythenon-exchangeenergiesarerequired(E(1),E(2) andE(2) )andtheALDAXkernelisused,thelimits elst ind,pol disp,pol aresignificantlyhigher. For example, if we impose a 16GB limit to the sizes of individual arrays that need to be stored completely in memory,thenthefollowinglimitsapply: • Mainbasissize:N =n +n wheren n ≤131072 o v o v Examples: ◦ N-methylpropanamide:n =24,thereforeN ≤5461. o ◦ Carbamazepine:n =62,thereforeN ≤2114. o ◦ C :n =180,soN ≤728. 60 o • Auxiliarybasissize:M ≤131072.Inpracticethelimitwillbelowerforcomputationalreasons. • Computationalbottlenecks: ◦ Hessians:CalculationofthehybridHessians(ALDA+CHF)isanO(n3n3)process.Thiscanbereduced o v using the methods recently developed by Bukowski et al. [2005]. This has been done for the ALDAX andALDAkernelswhichcanbecalculatedinO(M2n n )effortandwithO(M2)memoryrequirements. o v ◦ 2-electron 4-index integrals: These are required for the exchange energies. In particular the second- orderexchangeenergies. • Otherlimitations ◦ Notparallelised. ◦ Whilereasonablylargecalculationsarepossiblewiththisversionofthecode(seetheexamplesgiven above),bearinmindthatthisstillisaserialcodeandthatcalculationsofthesecond-orderexchange energies for large systems will be difficult as they will typically require more resources and a re- programmingoftherelevantCamCASP modulestomakebetteruseofmemoryanddiskresources. 4 Installation The CamCASP package comprises several different programs, and also uses a number of third-party programs — specifically, an ab initio package for the DFT or wavefunction calculations and the Orient program. While we have interfaced CamCASP to the DALTON, NWChem 6.x, GAMESS(US) and Psi4 programs, the scripts suppliedwithCamCASP supportPsi4,DALTONandNWChem6.xmorefullythantheothers.Consequently,we recommendthatthePsi4,DALTON(2.0,or2015orlater)orNWChem6.xabinitiopackagesareused,andthis 5 isassumedbelow.SeveralPythonscriptsareprovidedtosimplifythetaskofsettingupthevariousfilesthatare neededforthecalculation,andfeedingthemtotheprogramsthatdothework.ThescriptsrequirePython3.6or later.Inorderforthistoworksuccessfully,afewconventionsneedtobefollowedinthewaythatyourcomputeris organised,andbecauseoftheuseofthird-partyprogramstheinstallationprocedurecannotbeasfullyautomated asyoumightwish.ThefollowinginstructionsapplytoUnixandLinuxsystems;theycanbefollowedwithlittle changeforMacOSXandWindowssystemsbyworkingfromaterminalwindowontheMacorbyusingaUnix emulatorsuchasCygwinunderWindows. 1. TherecommendedprocedureistoclonethepackagefromtheCamCASP repository,byusing git clone https://gitlab.com/anthonyjs/camcasp.git directory Thiswilldownloadallthefilesintothedirectoryyouspecified.ItincludesprogrambinariesforLinux.Fora MacOSversion,usethesamecommandwiththeadditionaloption--branch CamCASP-6.0-macos 2. Next,youneedtoaddtheCamCASP /bindirectorytoyourpath;forexample: export PATH=/home/<user>/camcasp/bin:$PATH Thiscanbeincludedinyourinitializationfile.Thiswouldprobablybeyour.bashrcfileifyouusetheBash shell. 3. InstallOrient,mosteasilyusing git clone [email protected]:anthonyjs/orient.git directory TheOrientpackageincludesthesourcefiles,soyoucancompileityourself,butthereareexecutablebinaries includedforLinuxandOSX. 4. Install an ab initio wavefunction code – DALTON, NWChem 6.x or Psi4 – if you don’t already have one or more of them already installed. DALTON 2.0 and DALTON-2015 (and later) require slightly different input, butCamCASP canhandleeither.TheCamCASP interfacetoPsi4 isnewinversion6.0andshouldbetreated asexperimental.ThisinterfacehasbeentestedonthePsi4 versions1.1,1.2and1.3;thelattertwoversionsare availableontheGitHubpageofPsi4 andmustbebuiltbytheUser. 5. YoucouldalsoinstalltheGAMESS(US) program,butbearinmindthatforthepresent,wedonotincludeany examplesofscriptstoperformcalculationswithGAMESS(US) andCamCASP. 6. Dalton-specific:IfyouwishtousetheSapt2008programwiththeDALTON2.0program,itisnecessarytoap- plythepatchtoDALTON2.0includedinthedistribution.ThisallowscalculationsusingtheALDA/ALDA+CHF kernelswithxc-kernelintegralsdirectlyfromDALTONandallowstheFermi–Amaldi(FA)asymptoticcorrec- tion with the Tozer & Handy splicing scheme [Tozer and Handy, 1998]. The patch for DALTON needs to be applied,andDALTONrecompiledifnecessary.Forsourcesfortheseprogramssee http://www.kjemi.uio.no/software/dalton/dalton.html http://www-stone.ch.cam.ac.uk/programs.html#Orient4 http://www.physics.udel.edu/~szalewic/SAPT/SAPT.html DALTON2015doesn’trequirethepatch,asithastheasymptoticcorrectionalreadyincluded,andCamCASP cannowhandletheALDAandALDA+CHFkernelitself. 7. NWChem6.x-specific:TheCS00asymptoticcorrection[CasidaandSalahub,2000]willbeusedwithNWChem 6.x.Ideallyanenergyshiftshouldbeprovidedforthismethod;itisthesumofthe(positive)ionizationenergy and the (negative) HOMO eigenvalue. If this is not provided an empirical relationship between the HOMO eigenvalueandtheIPisusedinstead.Neitherisasgoodanapproximationastheasymptoticcorrectionscheme usedwithDALTON(above).TheCS00schemeissimilartotheGRACschemeofGruningetal.[2001]. 8. Psi4 -specific: CamCASP uses the main basis sets from the Psi4 basis library, so many more basis sets are possible. However you need to ensure that the relevant auxiliary basis is available in the CamCASP basis library.Psi4 usestheGRACasymptoticcorrection. 9. Windows specific: the interface codes that convert the results of the wavefunction calculation into a suitable form for CamCASP are provided as Linux or Mac OS binaries, and may need to be recompiled. To do this, changetotheCamCASPdirectoryandrun make interfaces Youwillneedagfortrancompiler. CamCASPhasbeentestedbutnotextensivelyusedbyusunderMacOSandnotusedatallunderWindows,so wecannotguaranteethatitwillrunsmoothlyundertheseoperatingsystems. 6 10. Nowyouneedtodefinesomeenvironmentvariables. • CAMCASP shouldbesettothefullpathnameofthebasedirectoryfortheCamCASP package. • SCRATCH shouldbesettothefullpathnameofascratchdirectorythatcanbeusedfortemporaryfiles. • ARCH shouldbeset,forhistoricalreasons,to’x86-64’ifyouarerunningunderLinuxandto’osx’if runningunderDarwin(MacOS). Thesevariablescanallbesetinyourinitializationscript,sothatyoudon’tneedtosetthemeverysession.For example: export CAMCASP=/usr/local/camcasp-6.0 YoumayalsoalsoneedtoaddtheSapt2008\bindirectorytoyourpathifyouplantousethisprogram: export PATH=$SAPT/bin:$PATH 11. If you wish to use NWChem 6.x or Psi4 on a multi-processor or multi-threaded machine, you may need to provide shell scripts nwchem.sh or psi4.sh in the CamCASP bin directory. They should deal with any initial setup and execute the command required to launch NWChem 6.x or Psi4 with the required number of processors. The procedure for doing this will vary from one computer to another, and we cannot provide a generalsolution,butthefilesnwchem.sh.exampleandpsi4.sh.exampleareprovidedasaguide. 12. Finally you need to ensure that CamCASP can access the SCF codes and other programs that it uses. These include at least one of the SCF codes Dalton (2013 or later), Dalton2006 (Dalton 2.0), NWChem and Psi4, andalsotheOrientprogramandpossiblytheSAPTprogram(i.e.theHartree-FockbasedSAPT–CamCASP providesSAPT(DFT)itself).ThisjustrequiresthattheexecutablefilescanbefoundinyourPATH.Sincethe CamCASP /bin directory must be in your path anyway, one way to achieve this is to set up symbolic links (aliases) in the CamCASP bin directory to these programs. This only needs to be done once. Alternatively it may be necessary, or just more convenient, to install a little shell script, e.g. psi4.sh, in the CamCASP /bin directorytoinvoketheprogram,asexplainedabove. TocheckthattheprogramsareallaccessibletoCamCASP,runthecommand setup.py (afterchangingthePATHandsettingtheenvironmentvariablesasexplainedabove). 13. Testingtheinstallation Onceeverythingissetup,youcanrunsometeststhatwillexercisetheprograms.GototheCamCASP subdi- rectorytestsandfollowtheinstructionsintheREADMEfile.Therearetestsforsingle-configurationdimer energycalculationsandforproperties.andforenergyscansoverasetofdimerconfigurations. 4.1 BuildingCamCASP fromsource ExecutablebinariesoftheCamCASP,Process,Casimir,PFITandClusterprogramsareprovidedforPCsrunning LinuxandforMacOSX.Wherepossiblethesearestaticbinariesanddon’trequireadditionallibraries.Ifyouneed tocompilethepackageyourself,youwillneedtoobtainaccesstothesource.Thisisnotnormallyavailable,but may be permitted on request. The Makefile supplied with the CamCASP source files will build the CamCASP , Process,Casimir,PFITandClusterprograms.Herearesomebriefinstructions. Thebuildwillhappeninadirectory$CAMCASP/arch/compilerwherearchisthemachinearchitectureandcompiler isthenameofthecompilerchosen.Thearchitectureandcompilermustbespecified,eitheronthemakecommand line: make all ARCH=archCOMPILER=compiler bychangingthebeginningoftheMakefiletosetthedefaultvalues,orbysettingenvironmentvariablesARCHand COMPILER. Thecompilerflagsusedaresetinthefile$CAMCASP/arch/compiler/exe/Flags.Youwillhavetosetmachine- specificflagshere.TheMACHINEvariableprovidesanalternativewaytoselectflagsfordifferentplatforms,using the same Flags file — see the examples provided. On a Linux machine the name of the machine should be de- terminedautomatically,butifthisdoesn’tworkyoucanspecifyitonthemakecommandline.Probablytheonly thingthatwillneedchangingistheLIBSflagthatsetsthelibrariestobelinkedto. Architectureoptionsarex86-64andosx. Compileroptionsare:pgf90,gfortran,andifort. 7