The MultiConfiguration SCF Program

Peter Knowles, Hans-Joachim Werner,

Cambridge University, 1984


Contents


Introduction


    THIS MANUAL DESCRIBES THE USE OF THE PROGRAM MULTI WITHIN
 THE ATMOL PROGRAM SYSTEM IN ORDER TO OBTAIN MCSCF WAVEFUNCTIONS. 
 THE BASIC METHOD IS THAT DESCRIBED IN REFERENCES 1 AND 2 
 (SEE REFERENCES), BUT STANDARD NEWTON-RAPHSON
 AND STEP-RESTRICTED AUGMENTED HESSIAN METHODS ARE PROVIDED AS
 ALTERNATIVES. THE CONFIGURATIONS ARE GENERATED EITHER IN A CSF
 BASIS BY EXCITATION LEVEL FROM AN ARBITRARY NUMBER OF
 REFERENCES, OR, FOR COMPLETE ACTIVE SPACE MCSCF WAVEFUNCTIONS,
 A DETERMINANT BASIS IS USED (THE LATTER ALLOWS VERY LONG CI
 EXPANSIONS TO BE HANDLED).


References


  1. H.-J. Werner and P.J. Knowles, J. Chem. Phys. 82 (1985) 5053.
  2. P.J. Knowles AND H.-J. Werner, Chem. Phys. Letters, 115 (1985) 259.
  3. P.J. Knowles AND N.C. Handy, Chem. Phys. Letters 111 (1984) 315.
  4. M.R. Hoffmann, D.J. Fox, J.F. W, Y. Osamura, Y. Yamaguchi, R.S. Grev, G. Fitzgerald, H.F. Schaefer III, P.J. Knowles and N.C. Handy, J. Chem. Phys. 80 (1984) 2660.
  5. B.R. Brooks, J. Chem. Phys.
  6. V.R. Saunders and J.H. van Lenthe "The Direct CI method, A detailed analysis" Mol. Phys. 48 (1983) 923-954.

The general structure of the program



   THE PROGRAM IS DIVIDED INTO 6 LOGICALLY SEPARATE MODULES:

START   - DATA INPUT AND INITIALISATION
SORT    - SORTS THE INTEGRAL TAPE INTO COULOMB OPERATORS IN A
          FORMAT SUITABLE FOR INTEGRAL TRANSFORMATION. AFTER
          THIS STAGE, THE INTEGRAL TAPE IS NO LONGER REQUIRED,
          AND SO ANY RESTART, FOR EXAMPLE, MAY BYPASS THIS
          STAGE SO LONG AS THE SORTED INTEGRAL FILE HAS BEEN
          SAVED.
FMTP    - CONSTRUCTS THE CI COUPLING COEFFICIENTS AND WRITES
          THEM TO THE CI FORMULA FILE. THIS IS A MODIFICATION
          OF THE GRAPHICAL UNITARY GROUP PROGRAM OF DR. B.R.
          BROOKS  5 . AGAIN, THIS STAGE MAY BE OMITTED IN A
          RESTART, PROVIDED THE FORMULA FILE HAS BEEN
          RETAINED. FOR CASSCF CALCULATIONS, THIS STAGE IS NOT
          REQUIRED, AND IS AUTOMATICALLY BYPASSED.
REDUND  - READS THE CI FORMULA FILE, AND DECIDES WHICH ACTIVE-
          ACTIVE ORBITAL ROTATIONS ARE REDUNDANT, USING THE
          METHOD OF REFERENCE 4.
WVFN    - CALCULATES THE MCSCF WAVEFUNCTION ITERATIVELY. THIS
          IS THE MAIN SECTION OF THE PROGRAM. EACH ITERATION
          CONSISTS OF TWO PARTS, INTEGRAL TRANSFORMATION (BY
          THE METHOD OF SAUNDERS AND VAN LENTHE, REFERENCE 6)
          AND WAVEFUNCTION OPTIMISATION. IN ADDITION, ONE MAY
          OPTIONALLY DIAGONALISE THE HAMILTONIAN MATRIX
          BETWEEN THESE TWO STAGES (FOR EXAMPLE IN THE FIRST
          ITERATION THIS IS USUALLY APPROPRIATE), OR PERFORM A
          MORE GENERAL OPTIMISATION OF CI COEFFICIENTS AND
          INTERNAL ORBITALS. THE WAVEFUNCTION OPTIMISATION MAY
          BE PERFORMED ON JUST THE ORBITALS ALONE, OR FOR THE
          ORBITAL AND CI VARIATIONS TAKEN TOGETHER, AND AT
          PRESENT FOR EITHER OF THESE POSSIBILITIES, THERE ARE
          THREE DIFFERENT OPTIMISATION METHODS AVAILABLE. SOME
          OF THESE METHODS USE THE HISTORY OF PREVIOUS
          ITERATIONS; ALL THIS INFORMATION IS PRESERVED ON THE
          DUMPFILE, AND SO THE PROGRAM MAY BE RESTARTED AT THE
          BEGINNING OF ANY ITERATION. THERE IS NO PROVISION
          FOR DUMPING AND RESTARTING IN THE MIDDLE OF AN
          ITERATION.
ANAL    - PERFORMS THE ANALYSIS OF THE CONVERGED WAVEFUNCTION.
          THERE IS AN ABSOLUTE LIMIT OF 255 BASIS FUNCTIONS, 31
          ACTIVE ORBITALS, AND 32767 CONFIGURATIONS (EXCEPT FOR 
          CASSCF WAVEFUNCTIONS).


Files used by the program



FORTRAN 5    CARD IMAGE INPUT
FORTRAN 6    PRINTED OUTPUT.
ATMOL   ED0  CI FORMULA FILE (NOT USED FOR CASSCF
             CALCULATIONS).
ATMOL   ED2  ATMOL INTEGRAL FILE. THIS MUST BE PROVIDED IF THE
             SORT STAGE IS TO BE EXECUTED, AND HOLDS THE TWO
             ELECTRON INTEGRALS OVER BASIS FUNCTIONS. IF
             MOLECULAR SYMMETRY IS TO BE USED, THEN THESE MUST
             BE SYMMETRY ADAPTED INTEGRALS. (IN THE MOLPRO
             VERSION, THE MOLPRO INTEGRAL FILE, RATHER THAN
             ED2, IS READ).
ATMOL   ED3  STANDARD ATMOL DUMPFILE. THIS PROGRAM WRITES ITS
             DUMP TO SECTION 195 BY DEFAULT. IF SECTION 190 IS
             PRESENT (I.E. A SYMMETRY ADAPTION HAS BEEN
             PERFORMED), THEN THE PROGRAM WILL USE THE
             SYMMETRY SPECIFIED THEREIN, AND EXPECTS SYMMETRY
             ADAPTED INTEGRALS.
ATMOL   ED1  SORTED INTEGRAL FILE. THIS IS CREATED BY THE SORT
             MODULE, AND SHOULD BE RETAINED IF A RESTART IS
             ENVISAGED.
ATMOL   ED4
        ED6
        ED7  SCRATCH FILES.

   THE FILES DESIGNATED AS ED0, ED2, ED3, ED1, ED4, ED6 ABOVE
 MAY BE REASSIGNED TO OTHER STREAM NUMBERS AS DESCRIBED BELOW.


The program directives


    THE OPTIONS ARE PROVIDED BY A SEQUENCE OF DIRECTIVES ON
 FORTRAN STREAM 5. A DIRECTIVE CONSISTS OF ONE OR MORE LINES OF
 INPUT BEGINNING WITH A KEYWORD, WHICH MAY BE TRUNCATED TO 8
 CHARACTERS. BETWEEN DIRECTIVES, THE DATA MAY CONTAIN ANY
 NUMBER OF COMPLETELY BLANK LINES, OR LINES HAVING A '*' IN
 COLUMN 1. THESE LINES ARE IGNORED, AND MAY BE INTRODUCED INTO
 THE DATA FOR CLARITY. IN ADDITION, THE FIRST LINE OF THE DATA
 FILE IS NOT SCANNED FOR DIRECTIVES, BUT RATHER ITS CONTENTS
 ARE TAKEN AS A TITLE FOR THE RUN. NEARLY ALL THE INFORMATION
 GIVEN IN DIRECTIVES IS STORED IN THE DUMP (WHICH DEFAULTS TO
 SECTION 195 OF THE DUMPFILE), SO IT IS NOT USUALLY NECESSARY
 TO RESPECIFY OPTIONS IN A RESTART JOB.

    IN THE DESCRIPTION OF THE DIRECTIVES WHICH FOLLOWS, THE
 STRUCTURE OF A COMMAND LINE IS SHOWN IN THE FORM

     KEYWORD A < B > < C < D > >

 THE UPPER CASE ITEM KEYWORD MUST APPEAR IN THE DATA AS SUCH.
 THE LOWER CASE ITEMS REPRESENT REAL, INTEGER OR CHARACTER
 QUANTITIES SPECIFIED BY THE USER. IN THIS EXAMPLE, 'A' MUST BE
 SUPPLIED, 'B' AND 'C' ARE OPTIONAL; 'D' IS OPTIONAL, BUT MAY
 BE SUPPLIED ONLY IF 'C' IS GIVEN. PARAMETERS A,B,C,.. MAY BE
 SEPARATED FROM EACH OTHER BY ONE OR MORE SPACES, OR A COMMA.
 MULTIPLE COMMAS MAY BE USED IN AN OBVIOUS WAY TO OMIT
 SPECIFICATION OF PARTICULAR OPTIONAL PARAMETERS (THEN DEFAULT
 VALUES ARE USUALLY USED). A SEMICOLON MAY OPTIONALLY BE USED
 TO TERMINATE A LINE (E.G. TO PREVENT FURTHER READING OF
 PARAMETERS, OR TO INTRODUCE COMMENTS AT THE END OF THE LINE).
 FLOATING POINT QUANTITIES MAY BE INPUT IN ANY OF THE FORMS 50
 50.0 5D1 5.0D1 5E1 5.0E1.


The first two lines of input

The BASIS or SYMMETRY directive

The first directive in a MCSCF job should always be a BASIS directive or a SYMMETRY directive.

The BASIS directive is meant to specify the number of basis functions in the calculation. The syntax is:

  BASIS NBASIS
where BASIS is a literal string and NBASIS is the number of basis functions.

The SYMMETRY directive specifies that the basis functions are symmetry adapted. In this case the program can extract the number of basis functions from the dumpfile. The syntax is:

  SYMMETRY
where SYMMETRY is a literal string.

NOTE: The first line of data after the BASIS or SYMMETRY directive is taken to be the TITLE of the job.


Directives for changing the default file specifications

The DUMPFILE directive

 DUMPFILE DD3 < IBLK3 < ISEC > >
SPECIFIES THAT THE DUMPFILE IS ON ATMOL FILE 'DD3' STARTING AT BLOCK 'IBLK3' (DEFAULT 1) WITH ONE ELECTRON INTEGRALS AT SECTION ISEC (DEFAULT 192). IN THE MOLPRO VERSION, THE DUMPFILE ED3 IS REQUIRED AS A SCRATCH FILE. THE DUMP CAN BE WRITTEN TO THIS FILE, OR TO ANY ALLOCATED MOLPRO FILE/SECTION.

NOTE: IF PRESENTED, THE DUMPFILE DIRECTIVE MUST BE THE FIRST DIRECTIVE AFTER THE JOB TITLE.

 DEFAULT: DUMPFILE ED3 1 192
The MAINFILE directive

 MAINFILE DDA

    SPECIFIES THAT THE TWO ELECTRON INTEGRALS AS PRODUCED BY
 THE INTEGRALS PROGRAM OR THE INTEGRAL TRANSFORMATION PROGRAM
 ARE TO BE READ FROM ATMOL FILE 'DDA' STARTING AT BLOCK 1 UNTIL
 END OF FILE.

 DEFAULT: MAINFILE ED2

The FILES directive

 FILES < DD2 < DD4 < DD6 < DDFT > > > >

    SPECIFIES THE ATMOL FILENAMES FOR THOSE FILES CREATED
 EXCLUSIVELY BY THIS PROGRAM. 'DD2' IS THE SORTED INTEGRAL
 FILE, 'DDFT' CONTAINES THE HAMILTONIAN FORMULAE, AND 'DD4' AND
 'DD6' ARE OTHER WORK FILES.

 DEFAULT: FILES ED1 ED4 ED6 ED0


Directives to specify the wavefunction


The ORBITAL directive

 ORBITAL
 < SPEC1 > < SPEC2 > ...
 ..
 ..
 END
 <  < RSPEC1 > < RSPEC2 > ...
 ..
 ..
 END  >

    THIS DIRECTIVE SPECIFIES MOST OF THE DETAILS OF THE MCSCF
 WAVEFUNCTION - THE NUMBER AND SYMMETRIES OF CORE AND ACTIVE
 ORBITALS, AND THE REFERENCE CONFIGURATION(S). THE FIRST LINE
 CONSISTS OF THE KEYWORD ORBITAL ALONE. THERE THEN FOLLOWS ONE
 OR MORE SUBSEQUENT LINES CONTAINING ONE WORD FOR EACH OCCUPIED
 ORBITAL, FINISHING WITH THE WORD END. THE DATA MAY BE SPREAD
 OVER AS MANY LINES AS IS CONVENIENT. THE FORMAT OF THE OBITAL
 DESCRIPTOR CODE IS

     < REPCNT > < KEY > CODE SYMMETRY

 WITH NO CONTAINED BLANKS. 'REPCNT' IS AN INTEGER GIVING THE
 NUMBER OF ORBITALS OF THIS SORT (DEFAULT 1). 'CODE' CONSISTS
 OF THREE LETTERS, AND MAY BE ONE OF

 FZC   THIS ORBITAL IS A FROZEN CORE ORBITAL (DOUBLY OCCUPIED
       IN ALL CONFIGURATIONS), I.E. ITS COEFFICIENTS ARE
       OMITTED FROM THE OPTIMISATION PROCEDURE, AND THE FINAL
       ORBITAL IS THE SAME AS SUPPLIED IN THE INITIAL GUESS.
 COR   THIS ORBITAL IS A CORE ORBITAL (DOUBLY OCCUPIED IN ALL
       CONFIGURATIONS).
 DOC   THIS ORBITAL IS DOUBLY OCCUPIED IN THE PRINCIPLE
       REFERENCE CONFIGURATION.
 ALP   THIS ORBITAL IS SINGLY OCCUPIED WITH ALPHA SPIN IN THE
       PRINCIPLE REFERENCE CONFIGURATION.
 BET   THIS ORBITAL IS SINGLY OCCUPIED WITH BETA SPIN IN THE
       PRINCIPLE REFERENCE CONFIGURATION.
 UOC   THIS ORBITAL IS UNOCCUPIED IN THE PRINCIPLE REFERENCE.

    ALL THE COR ORBITALS MUST BE SPECIFIED BEFORE THE FIRST
 ACTIVE ORBITAL, AND THERE MUST NOT BE MORE BET ORBITALS THAN
 ALP ORBITALS. NOTE THAT IF AN OPEN SHELL REFERENCE
 CONFIGURATION IS SPECIFIED, ALL POSSIBLE SPIN COUPLINGS ARE
 INCLUDED. THE TOTAL SPIN OF THE WAVEFUNCTION IS TAKEN TO BE
 THE RESULTING VALUE OF SZ FOR THE PRINCIPLE REFERENCE
 CONFIGURATION. 'SYMMETRY' IS A DIGIT FROM 1 TO 8 AND GIVES THE
 SYMMETRY OF THE ORBITAL UNDER THE POINT GROUP SYMMETRY
 ACTUALLY BEING USED. 'KEY' IF PRESENT SHOULD BE ONE OF / OR :.
 / INDICATES THAT SEVERAL REFERENCE CONFIGURATIONS ARE TO BE
 USED, AND THAT, IN THE OTHER REFERENCES, THIS ORBITAL MAY HAVE
 DIFFERENT OCCUPANCY. : INDICATES THAT THIS ORBITAL IS A
 'VALENCE' ORBITAL FOR THE PURPOSE OF GENERATING
 CONFIGURATIONS; IT IS POSSIBLE TO GENERATE ALL CONFIGURATIONS
 DIFFERING BY AT MOST N REPLACEMENTS WITHIN THE : ORBITALS, AND
 AT MOST M < N REPLACEMENTS WITHIN THE COMPLETE SET OF ACTIVE
 ORBITALS. THUS, FOR EXAMPLE, ONE MAY PRODUCE A WAVEFUNCTION
 WHICH IS "ALL SINGLES AND DOUBLES OUT OF A FULL CI VALENCE". :
 AND / MUST NOT BE MIXED IN THE SAME CALCULATION.

    IF NO ORBITALS HAVE BEEN FLAGGED AS / , THE DIRECTIVE IS
 TERMINATED BY THE END WORD, AND CORE AND ACTIVE ORBITALS ARE
 DEFINED FOR THE JOB. AT THIS POINT, THE PROGRAM ASSIGNS AND
 CHECKS SYMMETRIES FOR THE VIRTUAL ORBITALS, SYMMETRY BLOCKING
 THE VIRTUAL SPACE (AS WOULD BE REQUIRED FOR A SUBSEQUENT RUN
 OF THE ATMOL DIRECT CI PROGRAM). IF A / ORBITAL WAS IN THE
 LIST, FURTHER WORDS ARE READ IN DEFINING THE OTHER REFERENCE
 CONFIGURATIONS. ONLY THOSE ORBITALS WHICH WERE ORIGINALLY
 MARKED / SHOULD BE SPECIFIED, WITH THEIR OCCUPANCY AND THE
 CORRECT SYMMETRY (BUT NO / ). THE PROGRAM READS THE EXTRA
 REFERENCES ONE BY ONE UNTIL ANOTHER END IS REACHED, AND
 CLEARLY THE NUMBER OF CODES GIVEN SHOULD BE A MULTIPLE OF THE
 NUMBER OF / ORBITALS.

 DEFAULT: NONE; THE DIRECTIVE MUST BE GIVEN UNLESS THE RESTORE
 DIRECTIVE IS USED.
 
The EXCITATION directive

 EXCITATION < IEXC > < IEXCV >

    ALL CONFIGURATIONS DIFFERING IN AT MOST IEXC ACTIVE
 ORBITALS PLUS AT MOST IEXCV ':' ORBITALS COMPARED TO ANY
 REFERENCE CONFIGURATION ARE INCLUDED IN THE CONFIGURATION
 LIST.

 DEFAULT: EXCITATION 99999 (I.E. COMPLETE ACTIVE SPACE CI)
 UNLESS ANY ORBITAL HAS BEEN FLAGGED '/', WHEN THE DEFAULT IS
 EXCITATION 0.
 
 SPECIAL NOTE FOR CASSCF WAVEFUNCTIONS:

    IF A CASSCF CALCULATION IS BEING PERFORMED (NO EXCITATION,
 NO "/" ORBITALS), THEN THE DETERMINANT CI METHOD DESCRIBED IN
 REFERENCE 3 IS USED INSTEAD OF A FORMULA TAPE THROUGHOUT THE
 PROGRAM. A "DONT FMTP" DIRECTIVE IS ISSUED INTERNALLY TO
 PREVENT EXECUTION OF THE UNITARY GROUP FORMULA TAPE ROUTINES.
 AT THE TIME OF WRITING, IT IS NOT POSSIBLE TO USE THE
 DETERMINANT CI METHOD IN FULLY COUPLED NEWTON-RAPHSON OR
 AUGMENT CALCULATIONS; THEREFORE IN SUCH CASES, ONE SHOULD CODE
 "EXCITATION 1000" TO FORCE USE OF THE UNITARY GROUP CI RATHER
 THAN DETERMINANTS.
 
The STATE directive

 STATE NSTATE

    NSTATE ROOTS OF THE HAMILTONIAN ARE TO BE OPTIMISED
 SIMULTANEOUSLY, WITH WEIGHTS GIVEN BY THE WEIGHT DIRECTIVE.
 DEFAULT: STATE 1
 
The ROOT directive

 ROOT IROOT

    THE NSTATE CI VECTORS ON WHICH THE WAVEFUNCTION
 OPTIMISATION IS PERFORMED ARE TO BE TAKEN AS THE EIGENVECTORS
 IROOT, IROOT+1,.. IROOT+NSTATE-1 IN ASCENDING ENERGY ORDER.
 THIS ALLOWS A METHOD (WHICH IS NOT SAFE UNDER MANY CONDITIONS)
 OF OBTAINING AN EXCITED STATE WITHOUT HAVING TO OPTIMISE THE
 CI COEFFICIENTS OF THE LOWER STATES.

 DEFAULT: ROOT 1
 
The WEIGHT directive

 WEIGHT W1 W2 ...

    ASSIGN WEIGHTS TO THE NSTATE LOWEST HAMILTONIAN
 EIGENVECTORS, ENABLING A STATE-AVERAGED MCSCF CALCULATION TO
 BE PERFORMED. THIS ONLY WORKS PROPERLY FOR THE WERNER
 ITERATION PROCEDURE.

 DEFAULT: WEIGHT 0 0 ... 1


Directives to specify the trial wavefunction


The ORBGUESS directive

 ORBGUESS < ISEC > < CANONICAL > < DD > < IBLKD >

    TRIAL MOLECULAR ORBITALS WILL BE TAKEN FROM THE STANDARD
 TYPE=3 SECTION NUMBER 'ISEC', ON THE DUMPFILE ON FILE 'DD'
 STARTING AT BLOCK 'IBLKD'. IF 'DD' IS OMITTED THEN THE CURRENT
 DUMPFILE IS ASSUMED. IF 'IBLKD' IS OMITTED, IT IS ASSUMED TO
 BE 1. 'DD' AND 'IBLKD' HAVE NO EFFECT IF ISEC > 1000 (MOLPRO
 FILE). IF CANONICAL IS SPECIFIED, THE INITIAL ACTIVE ORBITALS
 ARE CANONICALISED IN THE FIRST ITERATION (ONLY FOR CASSCF
 WAVEFUNCTIONS). THIS MAY IMPROVE CONVERGENCE IN THE
 CONFIGURATION INTERACTION COEFFICIENTS, IN PARTICULAR IF THE
 INITIAL ORBITALS ARE THE NATURAL ORBITALS OF A PREVIOUS
 CALCULATION.

 DEFAULT: NONE; DIRECTIVE IS COMPULSARY UNLESS A RESTORE
 DIRECTIVE IS GIVEN.

The CIGUESS directive

 CIGUESS UNIT

 OR

 CIGUESS HERE NCI NSTATE
 ICI1 ISTATE1 VALUE1
 < ICI2 ISTATE2 VALUE2 >
 ..
 ..
 END

    SPECIFY THE INITIAL GUESS FOR THE CI VECTOR. IN THE SECOND
 FORM, THE NON-ZERO COEFFICIENTSOF THE CI VECTOR ARE TYPED IN
 EXPLICITLY. ONE OR MORE SUBSEQUENT LINES SPECIFY THE INDEX,
 STATE NUMBER, AND VALUE OF NON-ZERO COEFFICIENTS, THE
 DIRECTIVE BEING TERMINATED BY THE KEYWORD END. FOR NORMAL USE
 OF THE PROGRAM, THIS DIRECTIVE IS NOT USUALLY REQUIRED. IF
 HOWEVER ONE WISHES TO START WITH A SINGLE CONFIGURATION, BUT
 NOT DIAGONALISE THE CI MATRIX INITIALLY, THEN CIGUESS HERE
 MUST BE USED.
 
The ROTATE directive

 ROTATE ORB1.SYM1 ORB2.SYM2 < ANGLE >

    ROTATE INITIAL ORBITALS AMONGST EACH OTHER; ORB1.SYM1 IS
 THE ORB1'TH ORBITAL IN SYMMETRY SYM1. THE ANGLE DEFAULTS TO 90
 DEGREES (COMPLETE SWAP) IF OMITTED. ROTATE MUST BE GIVEN AFTER
 THE RESTORE OR ORBGUESS DIRECTIVE WHICH LOADS THE ORBITALS TO
 BE SWAPPED. IF SYM1, SYM2 ARE ZERO, THEN ORB1, ORB2 REFER TO
 ABSOLUTE ORBITAL NUMBERS.

 DEFAULT: NO ORBITALS SWAPPED.
 
The RESTORE directive

 RESTORE < ISEC < DD < IBLKD > > >

    RESTORE THE DUMP FROM A PREVIOUS RUN OF THE PROGRAM FROM
 SECTION 'ISEC' OF DUMPFILE STARTING AT BLOCK 'IBLKD' ON FILE
 'DD'. 'ISEC' DEFAULTS TO 195, 'DD','IBLKD' TO THE CURRENT
 DUMPFILE, AND, IF 'DD' IS SPECIFIED, 'IBLKD' TO 1. ONCE THE
 DUMP HAS BEEN READ IN, THE PROGRAM ATTEMPTS TO READ IN THE
 MOLECULAR ORBITALS TO WHICH THIS DUMP REFERS, ISSUING A
 ORBGUESS DIRECTIVE INTERNALLY. IN ADDITION, THE DUMPED CI
 VECTOR IS COPIED TO THE DUMP AS SPECIFIED ON THE START CARD,
 OTHERWISE UNCHANGED. HENCE THIS DIRECTIVE INITIATES A RESTART,
 RESETTING EVERYTHING AS IT WAS WHEN THE LAST RUN TERMINATED.
 IF THE OPTIONS FOR THE JOB ARE TO BE CHANGED, DIRECTIVES
 EFFECTING THIS MUST APPEAR AFTER THE RESTORE DIRECTIVE,
 OTHERWISE THE PARAMETERS THEY HAVE CHANGED WILL BE
 OVERWRITTEN. NOTE THAT OTHER INITIAL ORBITALS CAN BE LOADED
 WITH THE ORBGUESS DIRECTIVE.

 DEFAULT: NO RESTORE IS PERFORMED.


Directives controlling the wavefunction optimisation


The ITERATIONS directive

 ITERATIONS
 < DONT > < DO > KEY1 LIST1
 <  < DONT > < DO > KEY2 LIST2 >
 ..
 ..
 END

    THIS DIRECTIVE SPECIFIES THE METHOD OF OPTIMISATION TO BE
 USED IN EACH ITERATION, OVERRIDING THE DEFAULT ACTION. BETWEEN
 THE KEYWORDS ITERATIONS AND END ARE ONE OR MORE LINES
 CONTROLLING A PARTICULAR OPTION. THE PERMITTED OPTIONS WHICH
 ARE SELECTED BY 'KEY1', 'KEY2' ETC. ARE:

 DIAGCI   - DIAGONALISE THE HAMILTONIAN MATRIX AT THE START OF
            AN ITERATION.
 WERNER   - THE METHOD OF WERNER AND KNOWLES [1,2] TO BE USED.
 INTERNAL - AN INITIAL VARIATIONAL OPTIMISATION OF THE
            INTERNAL-INTERNAL ORBITAL ROTATIONS AND CI
            COEFFICIENTS TO BE PERFORMED AT THE START OF THE
            ITERATION, AS DESCRIBED IN    .
 NEWTON   - UNCONSTRAINED NEWTON-RAPHSON OPTIMISATION TO BE
            USED.
 AUGMENT  - AUGMENTED HESSIAN METHOD TO BE USED.
 UNCOUPLE - OPTIMISE ORBITALS AND CI COEFFICIENTS AS IF
            INDEPENDENT OF EACH OTHER.
 NULL     - NO OPTIMISATION IS PERFORMED.

    IF THE KEY APPEARS FIRST IN THE LINE OR IS PRECEDED BY THE
 KEYWORD DO, THE FLAG SELECTING THE OPTION IS TO BE SWITCHED
 ON, AND IF THE KEY IS PRECEDED BY DONT, IT IS TO BE SWITCHED
 OFF. AFTER 'KEY', A LIST OF ITERATION NUMBERS FOR WHICH THE
 FLAG IS TO BE SPECIFIED SHOULD FOLLOW. THIS MAY TAKE THE FORM
 OF A SIMPLE SERIES OF INTEGERS NOT GREATER THAN 40, OR MAY
 CONTAIN I TO J MEANING ALL ITERATIONS BETWEEN I AND J
 INCLUSIVE. ITERATION NUMBER 40 HAS A SPECIAL MEANING; THE
 OPTIONS APPLYING TO ITERATION 40 WILL BE USED AS SOON AS
 CONVERGENCE IS DETECTED. THUS, FOR EXAMPLE,

     NULL 40

 SUPPRESSES UPDATING OF THE WAVEFUNCTION ONCE CONVERGENCE IS
 REACHED. SOME OF THE OPTIONS ARE OBVIOUSLY MUTUALLY EXCLUSIVE
 (E.G. WERNER-AUGMENT), WHILST SOME OPTIONS IMPLY OTHERS; WHEN
 A 'DO' LINE IS PROCESSED, THESE CONSIDERATIONS ARE ENFORCED BY
 THE PROGRAM.
 
THE DEFAULT SETTINGS OF THESE OPTIONS IS GIVEN BELOW UNDER THE BEGIN DIRECTIVE. IT SHOULD NOT NORMALLY BE NECESSARY TO CHANGE FROM THE DEFAULT. DEFAULT: THE ITERATIONS INFO IS SET UP INITIALLY BY THE BEGIN DIRECTIVE. The DO and DONT directives

 DO   KEY1 < KEY2 > < KEY3 > ...
 DONT KEY2 < KEY2 > < KEY3 > ...

    THESE DIRECTIVES SPECIFY WHETHER OF NOT A PARTICULAR
 SECTION OF THE PROGRAM IS TO BE EXECUTED IN THIS RUN. 'KEY1'
 'KEY2' ... SHOULD BE ONE OF

     SORT      - PRE-SORT OF ATMOL INTEGRAL TAPE
     FMTP      - CONSTRUCTION OF HAMILTONIAN FORMULAE
     WVFN      - WAVEFUNCTION OPTIMISATION
     ANAL      - WAVEFUNCTION ANALYSIS
     CANONICAL - ORBITAL CANONICALISATION
     ACCOUNT   - CPU USAGE ACCOUNTING
     PRINT     - ADDITIONAL OUTPUT

    FOR EACH OF THESE FOUR MODULES THERE IS A FLAG STORED IN
 THE DUMP, WHICH IS INITIALLY ON. WHEN THE MODULE HAS COMPLETED
 ITS JOB, THE FLAG IS SWITCHED OFF. SPECIFYING DO FORCES THE
 FLAG TO BE SWITCHED ON, DONT FORCES IT TO BE SWITCHED OFF. FOR
 EXAMPLE, IF A START-UP CALCULATION IS BEING PERFORMED AND THE
 FORMULA TAPE FROM ANOTHER GEOMETRY IS AVAILABLE, THEN IT WOULD
 BE APPROPRIATE TO CODE

     DONT FMTP

 IF, ON THE OTHER HAND, ONE IS RESTARTING A CALCULATION IN
 WHICH THE SORT AND FMTP STAGES HAVE BEEN PASSED, BUT THE
 SORTED INTEGRAL FILE AND THE CI FORMULA FILE HAVE BEEN LOST,
 THEN ONE SHOULD CODES

     DO FMTP SORT

 AFTER THE RESTORE DIRECTIVE TO ENSURE THAT THESE STAGES ARE
 REPEATED IN THE RESTART JOB.

 DEFAULT: THE FLAGS ARE NOT ALTERED.
 
The CONVERGENCE directive

 CONVERGENCE CONV MODE

    A CONVERGENCE THRESHOLD IS RESET FROM ITS DEFAULT VALUE.
 'CONV' IS A FLOATING POINT NUMBER; 'MODE' IS A WORD WHICH
 SHOULD BE ONE OF ENERGY GRADIENT STEP (OR BLANK, WHEN ENERGY
 IS ASSUMED); THIS SPECIFIES WHICH CONVERGENCE PARAMETER IS
 CHANGED. THE PROGRAM ASSUMES THAT CONVERGENCE HAS BEEN REACHED
 WHEN ONE OF THE THREE CONVERGENCE CONDITIONS IS SATISFIED.
 THESE ARE: A) CHANGE IN ENERGY FROM LAST ITERATION, OR
 PROJECTED CHANGE TO NEXT ITERATION, LESS THAN THRESHOLD; B)
 GRADIENT OF ENERGY LESS THAN THRESHOLD; C) CHANGE MADE TO
 WAVEFUNCTION PARAMETERS LESS THAN THRESHOLD.

 DEFAULT: AS SET UP BY THE BEGIN DIRECTIVE

The MAXCYCLE directive

 MAXCYCLE MAXCYC

    RESTRICTS THE MAXIMUM NUMBER OF ITERATIONS TO 'MAXCYC'; IF
 CONVERGENCE IS NOT ACHIEVED WITHIN THIS LIMIT, THE PROGRAM
 TIDIES UP AND ENTERS THE ANALYSIS MODULE.

 DEFAULT: THE MAXIMUM NUMBER OF CYCLES IS SET INITIALLY BY THE
 BEGIN DIRECTIVE.
 
The STEP directive

 STEP < RADIUS < TRUST1 < FAC1 < TRUST2 < FAC2 > > > > >

    THIS DIRECTIVE RESETS THE INITIAL VALUES FOR THE TRUST
 RADIUS AND THE PARAMETERS WHICH ARE USED TO UPDATE IT, FOR
 AUGMENTED HESSIAN CALCULATIONS. 'RADIUS' SPECIFIES THE VALUE
 OF THE TRUST RADIUS FOR THE NEXT ITERATION; THE OPTIMISATION
 ALGORITHM COMPUTES A CHANGE IN THE WAVEFUNCTION WHOSE NORM IS
 RESTRICTED TO BE LESS THAN OR EQUAL TO 'RADIUS'. AT EACH
 ITERATION, THE PROGRAM COMPUTES WHAT IT THINKS THE ENERGY WILL
 BE AT THE NEXT ITERATION ON THE BASIS OF THE COMPUTED STEP IN
 THE WAVEFUNCTION, AND ASSUMING A SECOND ORDER ENERGY
 EXPRESSION. IN THE NEXT ITERATION, THE TRUST RATIO

       R  =  DELTA E (ACTUAL) / DELTA E (PREDICTED)

 IS COMPARED TO 1; IF THE DEVIATION IS > 'TRUST1', THEN
 'RADIUS' IS MULTIPLIED BY 'FAC1' (<1); IF THE DEVIATION IS <
 'TRUST2', THEN 'RADIUS' IS MULTIPLIED BY 'FAC2' (>1).

 DEFAULT: THE STEP PARAMETERS ARE SET UP INITIALLY BY THE BEGIN
 DIRECTIVE.
 
The COPT directive

 COPT COPVAR SELECT CISHFT CIACC ICIMAX ICIMX1 ICIMX2 MAXCI
 ICSTRT ICSTEP

    CONTROLS CONVERGENCE PARAMETERS IN OPTIMISATION OF CI
 COEFFICIENTS. 
The various parameters mean the following:
COPVAR
Start threshold for CI-optimization.
SELECT
Selection threshold for primary configurations (p-space). The configurations having an energy in the range E0 .. E0+SELECT are select for the primary space (E0 = groundstate). If a SELECT is less than zero all configuration will be selected.
CISHFT
Denominator shift for the q-space.
CIACC
Gradient threshold for CI diagonalisation.
ICIMAX
Maximum number of CI-optimizations in the first macro iteration.
ICIMX1
Maximum number of CI-optimizations in the second and subsequent iterations.
ICIMX2
Maximum number of CI-optimizations in the internal absorption step.
MAXCI
Maximum number of CI-optimizations per micro iteration.
ICSTRT
Number of micro iterations before the first CI optimization.
ICSTEP
Micro iteration increment between CI optimizations.
The default is:
 COPT 0.1 0.4 0.4 0.1   3 10 1 1   4 1
 
The NONLINEAR directive

 NONLINEAR ITMAXR IPRI DRMAX DRDAMP GFAK1 GFAK2 GFAK3 IRDAMP
 NTEXP

    THIS DIRECTIVE CONTROLS CONVERGENCE PARAMETERS IN THE
 MICROITERATIONS OF THE WERNER OPTIMISATION. ADVICE SHOULD BE
 SOUGHT BEFORE ATTEMPTING TO USE OTHER VALUES.
 
The THRESH directive

 THRESH VARMIN VARMAX

    CONTROLS CONVERGENCE THRESHOLDS IN WERNER OPTIMISATION.

The INTOPT directive

 INTOPT MAXITO MAXITC MAXREP NITREP IUPROD

    THIS DIRECTIVE CONTROLS THE WAY IN WHICH THE INTERNAL
 OPTIMISATION (INTERNAL) IS PERFORMED. ADVICE SHOULD BE SOUGHT
 BEFORE USE.
The program optimizes the orbitals and the CI coefficients simultaneously. In the optimization of the internal orbitals this leads to an outer loop containing the internal orbital optimization and CI optimization. The orbital optimization is contained in an inner loop. The parameters control:
MAXITO
The number of times the inner loop is executed in each pass through the outer loop.
MAXITC
The number of times the outer loop is executed in a internal optimization step.
MAXREP
The number of times the Werner optimization is repeated in one MCSCF iteration.
NITREP
The first NITREP-1 MCSCF iterations the Werner optimization is performed only once. If MAXREP greater than 1 from NITREP iterations on the Werner optimization will be performed MAXREP times each iteration.
IUPROD
Controls the generation of the orthogonal transformation matrix T. IUPROD may be set to 0, 1 or 2. By default the T=U-1=exp(R)-1 matrix is generated. If IUPROD is 1 then T=U(ext)*U(int)-1 else if IUPROD is 2 T=U(int)*U(ext)-1.
The default is:
 INTOPT 3 1 1 1 0
 
The DIIS directive

 DIIS DISVAR AUGVAR MAXDIS MAXAUG IDSCI IGWGT IGVEC IDSTRT
 IDSTEP

    THIS CONTROLS THE DIIS ACCELERATION OF MICROITERATIONS IN
 BOTH INTERNAL AND WERNER STAGES OF THE CALCULATION.
 
The SPARSITY directive

 SPARSITY SPARSE

    THIS OPTION IS RELEVANT ONLY FOR NEWTON OR AUGMENT
 OPTIMISERS. IT CONTROLS THE ENFORCED SPARSITY OF EXPANSION
 VECTORS IN THE LINEAR EQUATION SOLVER (SUCH SPARSITY IS USED
 TO REDUCE THE COMPUTATIONAL EFFORT). CHANGING THE DEFAULT IS
 NOT RECOMMENDED.

 DEFAULT: SPARSITY 1.0E-4
 
The BEGIN directive

 BEGIN

    THIS DIRECTIVE INITIALISES THE ITERATION CONTROLS FOR A
 STANDARD START-UP CALCULATION. IT IS ISSUED INTERNALLY BEFORE
 ALL OTHER DIRECTIVES, BUT IF THE RESTORE DIRECTIVE IS USED TO
 OVERWRITE THE DUMP, IT MAY BE DESIRABLE TO SUBSEQUENTLY ISSUE
 THIS DIRECTIVE EXPLICITLY, IF FOR EXAMPLE THIS IS A START-UP
 CALCULATION USING THE WAVEFUNCTION FROM A NEARBY GEOMETRY AS
 INITIAL GUESS. ALL INPUT PARAMETERS ARE THE RETAINED FROM THE
 PREVIOUS CALCULATION, BUT THE CALCULATION STARTS AT ITERATION
 NUMBER ONE AS A STARTUP JOB.

 DEFAULT: THE BEGIN DIRECTIVE IS ISSUED INTERNALLY BEFORE ALL
 ELSE.
 
The SAFETY directive

 SAFETY SAFE1 

    USED TO CONTROL CPU TIME SAFETY FACTORS WHICH DETERMINE
 WHEN A CALCULATION SHOULD BE DUMPED BECAUSE THERE IS NOT
 ENOUGH CPU TIME FOR A FURTHER ITERATION.

 DEFAULT: SAFETY 0.9 0.0

The IPRINT directive

 IPRINT   ...

    THIS DIRECTIVE CONTROLS WHAT IS PRINTED OUT AT EACH
 ITERATION. THE FORMAT IS THE SAME AS FOR THE PRINT DIRECTIVE
 DESCRIBED BELOW.

 DEFAULT: INITIALLY, ALL INTERMEDIATE PRINT OPTIONS ARE OFF.

The PSPACE directive

This directive allows the direct specification of the primary configuration space.

The syntax of this multiple line directive is

 PSPACE 
     ICONF
   < ICONF1 >
   < ICONF2 >
    ...
 END
where ICONF, ICONF1, and ICONF2 are integer configuration numbers.

If this directive is omitted a selection mechanism based on configuration energies will be used to define the primary space. This mechanism may be tailored through the COPT directive.


Directives controlling the analysis of the wavefunction


The PRINT directive

 PRINT < KEY1 >  < KEY2 >  ...

    CONTROLS PRINTING OF VARIOUS QUANTITIES. 'KEY1','KEY2',ETC.
 SHOULD BE CHOSEN FROM

     DRT      - PRINT DISTINCT ROW TABLE.
     FORMULAE - PRINT COMPLETE LIST OF ONE AND TWO ELECTRON
                FORMULAE.
     INTEGRAL - PRINT INTEGRALS OVER MOLECULAR ORBITALS.
     ADDRESS  - PRINT VARIOUS ADDRESSING ARRAYS.
     DEBUG    - PRINT DEBUGGING OUTPUT.
     DENSITY  - PRINT ONE AND TWO PARTICLE DENSITY MATRICES.
     HESSIAN  - PRINT HESSIAN MATRIX. THIS OPTION FORCES AN
                EVALUATION OF HESSIAN ON VECTOR FOR EACH ORBITAL
                ROTATION AND EACH CI COEFFICIENT, AND SO CAN BE
                EXTREMELY EXPENSIVE FOR LARGE PROBLEMS.
     GRADIENT - PRINT THE GRADIENT OF THE ENERGY WITH RESPECT TO
                WAVEFUNCTION CHANGES.
     ORBITALS - PRINT MOLECULAR ORBITAL COEFFICIENTS FOR THE
                OCCUPIED ORBITALS.
     VIRTUALS - PRINT MOLECULAR ORBITAL COEFFICIENTS FOR THE
                UNOCCUPIED ORBITALS.
     NATORB   - DETERMINE AND PRINT NATURAL ORBITALS.
     CIVECTOR - PRINT THE CI VECTOR.
     STEP     - PRINT THE COMPUTED CHANGE IN THE WAVEFUNCTION AT
                THE END OF AN ITERATION.

    FOR EACH OF THESE OPTIONS, THERE IS A FLAG INITALLY SET
 OFF. ON ENTRY TO THE PRINT DIRECTIVE, ALL THE FLAGS ARE FORCED
 OFF, AND THEN A KEYWORD WILL SWITCH ON ITS CORRESPONDING PRINT
 OPTION. THUS

     PRINT

 WILL SWITCH OFF ALL PRINT OPTIONS, WHILST

     PRINT DRT FORMULAE
     PRINT DENSITY

 WILL RESULT IN PRINTING OF THE DENSITY MATRIX ONLY. THE IPRINT
 DIRECTIVE HAS THE SAME FORM, BUT CONTROLS PRINTING IN EACH
 ITERATION OF THE WVFN MODULE. SOME OF THE FIELDS DESCRIBED
 ABOVE ARE NOT RELEVANT TO IPRINT, AND SOME ARE NOT RELEVANT TO
 PRINT.

 DEFAULT: INITIALLY, ALL PRINT OPTIONS ARE OFF.
 
The CANONICAL directive

 CANONICAL ISEC < CANI > < CANT > < CANA > < CI > < ISECO >

    THIS DIRECTIVE REQUESTS THAT THE CONVERGED MOLECULAR
 ORBITALS BE ROTATED SUCH THAT A SPECIFIED MATRIX IS
 DIAGONALISED. THESE CANONICALISED ORBITALS ARE WRITTEN TO
 SECTION ISEC OF THE DUMPFILE (ONLY IF ISEC IS NOT ZERO). THE
 MATRIX IS SPECIFIED BY THE THREE KEYWORDS CANI (CONTROLLING
 THE CANONICALISATION OF CORE ORBITALS), CANT (ACTIVE
 ORBITALS), CANA (VIRTUAL ORBITALS) WHICH SHOULD EACH BE SET TO
 ONE OF THE FOLLOWING TEXT STRINGS:

     NONE
     DENSITY
     FOCK

 IF CI IS SPECIFIED, THE HAMILTONIAN IS DIAGONALISED FOR THE
 NEW ORBITALS. IF ISECD IS NON-ZERO, THE JOB INFORMATION AND
 NEW CI COEFFICIENTS ARE WRITTEN TO THAT SECTION.
 
The PROPERTY directive

 PROPERTY PROP1 PROP2 PROP3 ... 

    THIS CAUSES THE CALCULATION OF ONE ELECTRON EXPECTATION VALUES. 
 PROP1, PROP2, ETC. SHOULD EACH BE ONE OF THE FOLLOWING WORDS

     T      - KINETIC ENERGY
     X      - X DIPOLE
     Y      - Y DIPOLE
     Z      - Z DIPOLE

 DEFAULT: NO PROPERTIES EVALUATED.


Terminating the data


The START directive

 START < ISECMO < ISECD > >

    THIS DIRECTIVE CAUSES COMPLETION OF THE EFFECTS OF SOME OF
 THE OTHER DIRECTIVES, AND CAUSES EXECUTION OF THE PROGRAM TO
 BEGIN. THE MOLECULAR ORBITALS ARE STORED ON SECTION 'ISECMO'
 OF THE DUMPFILE, AND THE DUMP IS TO SECTION 'ISECD'. 'ISECMO'
 DEFAULTS TO THE SECTION NUMBER FROM WHICH TRIAL MOS WHERE
 TAKEN. 'ISECD' DEFAULTS TO THE SECTION FROM WHICH A RESTORE
 WAS MADE, OR ELSE 195 IF NO RESTORE DIRECTIVE HAS APPEARED.
 ANY DIRECTIVES AFTER THIS ONE ARE COMPLETELY IGNORED BY THE
 PROGRAM.


The default values


 

    THE CONVERGENCE CONTROL VARIABLES WHICH CAN BE CHANGED AS
 DESCRIBED IN THE ABOVE SECTION ARE INITIALLY GIVEN THE
 FOLLOWING DEFAULT VALUES:

     MAXCYCLE 6
     DO WVFN ANAL
     CONVERGENCE 0.0 GRADIENT
     CONVERGENCE 0.0 STEP
     CONVERGENCE 1E-9 ENERGY
     ITERATIONS
     DIAGCI 1
     DONT DIAGCI 2 TO 40
     DONT NEWTON 1 TO 40
     DONT AUGMENT 1 TO 40
     DONT UNCOUPLE 1 TO 40
     DONT NULL 1 TO 39
     WERNER 1 TO 39
     INTERNAL 2 TO 39
     NULL 40
     END
     STEP 0.5 0.3 0.66 0.15 1.2
     COPT 0.1 0.4 0.4 0.1 3 10 1 5 1
     NONLINEAR 50 0 0.1 0.75 4.0 3.0 0.7 0 10
     INTOPT 3 1 1 1 0
     DIIS 0.07 0.2 10 10 2 0 1 2 1
     THRESH 1E-7 1E-3
     CANONICAL,,FOCK,,FOCK
 

Examples of complete data files


 

 A) A CALCULATION ON THE WATER MOLECULE. THE MCSCF CI
 EXPANSION IS COMPLETE WITHIN THE 'VALENCE' ORBITALS (I.E. A
 STANDARD CASSCF CALCULATION), AND CONVERGENCE CONTROLS
 APPROPRIATE FOR TRIAL SCF ORBITALS (WHICH ARE TAKEN FROM
 SECTION 2 OF THE DUMPFILE) ARE GIVEN.

 WATER -- MINIMAL VALENCE SPACE CASSCF
 ORBITAL
 FZC1 2DOC1 DOC4 DOC3 2UOC1 UOC4 END
 ORBGUESS 2
 START 5

    IF THE JOB RUNNING THE ABOVE DATA DOES NOT CONVERGE BEFORE
 CPU TIME RUNS OUT, AND THE SORTED INTEGRAL FILE IS SAVED, THEN
 THE FOLLOWING DATA WOULD BE APPROPRIATE FOR THE RESTART:

 WATER -- RESTART JOB
 RESTORE
 START
 
 B) A CALCULATION OF THE LOWEST SINGLET SIGMA STATE OF THE
 BEO MOLECULE (SEE BAUSCHLISCHER & YARKONY, J. CHEM. PHYS.
 72(1980) 1138).

 BEO -- LOWEST SIGMA STATE
 ORBITAL
 3COR1 /DOC1 /DOC3 /DOC4 /UOC1 /UOC3 /UOC4 END
        ALP1  DOC3  DOC4  BET1  UOC3  UOC4
        DOC1  DOC3  ALP4  UOC1  UOC3  BET4
        DOC1  ALP3  DOC4  UOC1  BET3  UOC4
        UOC1  DOC3  DOC4  DOC1  UOC3  UOC4
 END
 ITERATIONS
 UNCOUPLE 1 TO 3
 AUGMENT 1 TO 40
 END
 PRINT CIVECTOR
 CONVERGENCE 1.0E-10
 ORBGUESS 2
 START 5
 
 C) TO OBTAIN THE FIRST EXCITED SINGLET SIGMA STATE OF BEO,
 AFTER COMPLETION OF THE EXAMPLE (B) ABOVE, HAVING RETAINED ALL
 THE NECESSARY FILES

 BEO -- SECOND SINGLET SIGMA STATE
 RESTORE
 BEGIN
 STATE 2
 ITERATIONS
 AUGMENT 1 TO 40
 END
 CONVERGE 1E-10
 PRINT CIVECTOR ORBITALS
 DONT FMTP SORT
 START 10 100


Running the program


 

  CAMBRIDGE MVS, ATMOL

 LIBRARY ATMOL.ATMOL4.LINKLIB:ATMOL
 GLOBAL ED2=...  ED3=...  ETC.
 ATMOL ALLOC
 ATMOL INTEG :H:
 < DATA FOR INTEGRALS>
 :
 ATMOL TRAN4 :H:
 < DATA FOR SYMMETRY ADAPTION >
 :
 ATMOL SCF :H:
 < DATA FOR SCF >
 :
 ATMOL MULTI :H:
 < DATA FOR MCSCF >
 :

  CRAY, ATMOL
 
 ACCESS,DN=INTEGV,ID=KZCCHEM.
 ACCESS,DN=ADAPT,ID=KECA001.
 ACCESS,DN=SCF,ID=KZCCHEM.
 ACCESS,DN=MULTI,ID=KECA001.
 INTEGV.
 ADAPT.
 SCF.
 MULTI,CORE=300K.
 /EOF
 < DATA FOR INTEGRALS >
 < DATA FOR SYMMETRY ADAPTION >
 < DATA FOR SCF >
 < DATA FOR MCSCF >
 /EOJ

    N.B. 300K MEANS 300*1024 WORDS OF WORKSPACE, WHICH IS ABOUT
 WHAT IS AVAILABLE FOR A JOB SIZE OF 440000 WORDS.

  CRAY, MOLPRO

 ACCESS,DN=MULTIM,ID=KECA001.
 MULTIM,CORE=300K.
 /EOF
 < DATA FOR MCSCF >
 /EOJ

    IF INTEGRAL AND SCF CALCULATIONS ARE ALSO TO BE DONE, IT IS
 USUALLY CONVENIENT TO USE THE FOLLOWING INPUT SEQUENCE:

 ACCESS,DN=INT,ID=KECA001.
 ACCESS,DN=MULTIM,ID=KECA001.
 COPYF,O=D1.
 COPYF,O=D2.
 REWIND,DN=D1.
 ASSIGN,DN=D1,A=FT05.
 INT.
 REWIND,DN=D2.
 ASSIGN,DN=D2,A=FT05.
 MULTIM,CORE=300K.
 SAVE,DN=...,ID=...             (INTEGRAL FILE)
 /EOF
 < DATA FOR INTEGRALS + SCF >
 /EOF
 < DATA FOR MCSCF >
 /EOJ
 


Trouble Shooting


It is attempted to give some additional information to the error messages generated by the program. Quoted are error messages or parts of error messages, followed by descriptions, hints, etc.

"non zero p-space gradient"

This error may occur in internal optimisation. With this method the internal-internal orbital rotations and the CI coefficients are optimised simultaneously. This approach requires that the CI gradient is small otherwise erratic orbital rotations may result.

To avoid the error try switching off the internal optimisation using the iterations directive.

"large normalization factor in aughes"

This warning comes from the subroutine aughes (augmented hessian). The warning is printed in a loop that tries to orthogonalise a vector on a set of other vectors. If after the orthogonalisation a zero-vector remains the program is not able to normalise it properly, and it tries to orthogonalise it again. This may cause the program to dive into a tied loop.

To avoid this problem one should make sure that the program does not execute the aughes-subroutine. The can effected by switching off the internal optimisation and the werner optimisation, and switching on the uncoupled optimisation and the diagci option using the iterations directive.

"a not pos def in updating alpha"

This error usually occurs in the same situation a described under "large normalization factor in aughes". Often the latter warning precedes the error message. In such case the error results from a divide by zero in an attempt to normalise a zero-vector.

To avoid this problem one should make sure that the program does not execute the aughes-subroutine. The can effected by switching off the internal optimisation and the werner optimisation, and switching on the uncoupled optimisation and the diagci option using the iterations directive.

"no convergence"

Although this is message is not printed as an error message (it is just printed like a casual message), it means that the program found the calculation reached a hopeless state and gave up. This means that the status of all the information that is printed afterwards is undefined.

If the case in hand is well defined this problem may be circumvented by adjusting the CI-optimization parameters using the copt directive.


Glossary


P-SPACE
The space of so called primary configurations in the CI part of the MCSCF. These configurations are selected according to their energy, and are considered to be of primary importance to the state under consideration. In a sense the role of these configurations is similar to the role of the reference configurations in MRCI calculations. The selection criterion may be set using the COPT directive.
NOTE: There is no relation between this definition of p-space and momentum space, which sometimes physicists call p-space too.
Q-SPACE
The space of all configurations in the CI part of the MCSCF that are not part of the P-SPACE. The name q-space is due to the letter q following the letter p in the alphabet.
NOTE: There is no relation between this definition of q-space and real space, which sometimes physicists call q-space too.