GAMESS-UK USER'S GUIDE and REFERENCE MANUAL Version
Contents
1. HOH 109 05 HESSIAN 0 2 END BASIS TZVP RUNTYPE OPTIMIZE ED3 VECTORS GETQ ED3 1 1 ENTER 1 Let us assume the above calculation terminated on time up typical restart data is shown below Note that the RUNTYPE directive is as before i e still restoring the 3 21G Hessian If the ED3 specification is removed from this line the job would fail immediately since the starting Hessian will now be different from that resident on the history file DUMPFILE ED3 300 RESTART OPTIMIZE TITLE H20 TZVP RESTARTING FROM 3 21G ZMAT ANGSTROM 0 H 1 OH H 1 OH 2 HOH VARIABLES OH 0 956 HESSIAN 1 1 HOH 109 05 HESSIAN 0 2 END BASIS TZVP RUNTYPE OPTIMIZE ED3 VECTORS 1 ENTER 1 4 The optimisation appears to have converged but iterations continue if too small a value of XTOL is specified the energy is not sufficiently numerically accurate to justify using line searches This is especially true when employing default accuracy integrals 5 The optimisation will not converge e Check that the starting guess is reasonable e In a SADDLE calculation the quality of the initial hessian is crucial and the user should if possible generate this at the outset of the optimisation ie specify TYPE 3 for all the variables it should have only one negative eigenvalue if the starting point is reasonable e f a large step is made it is possible for the wavefunction to change state Usually this is manifested by the linear search retu2. 2 1481952 ANGS 0 699998 HCAH 181 0076613 DEGS 0 034888 ERROR ON Z MATRIX CARD NUMBER 3 ANGLE ALPHA IS OUTSIDE THE VALID RANGE OF O TO 180 4 TRANSITION STATE AND GEOMETRY OPTIMISATION 6 KHK HEK Hk HEK Hk HEK Hk HEK EAE AEA AEA EAE AERA EA HE Hk Ek Hk HE HE REAR AE A EAE RHE ERROR DETECTED FOO kk kk k ERROR DETECTED IN CONVERTING Z MATRIX TO CARTESIAN COORDINATES KHK HEK Hk HEK Hk HEK Hk HEK Hk HEK HEK HEK Hk HEK HEK HEK HE H HE Hk Ek Hk HE HE HE HE Hk H HE HHE In most cases this can be avoided by using the half angle and dummy centre specification as shown below TIME 60 TITLE CAH2 3 21G ZMAT ANGS CA X 1 1 0 H 1 CAH 2 XCAH H 1 CAH 2 XCAH 3 180 0 VARIABLES CAH 2 148 HESSIAN 0 7 XCAH 75 0 HESSIAN 0 1 END RUNTYPE OPTIMIZE VECTORS EXTGUESS ENTER 1 An additional constraint in the default optimisation algorithm is the inability to handle a change in point group during the optimisation pathway This example has been chosen to illustrate this point thus while the above data file avoids the angle problem it will again fail because of the point group change as shown below Z MATRIX ANGSTROMS AND DEGREES CD CENT ATOM N1 LENGTH N2 ALPHA N3 BETA 1 2 X 1 1 000000 1 3 2 H 1 2 148105 2 2 89 998 4 4 H 1 2 148105 3 2 89 998 5 3 180 000 6 VARIABLE VALUE HESSIAN CAH 2 1481048 ANGS 0 699995 XCAH 89 9978901 DEGS 0 156813 CHANGE IN POINT GROUP CNV 2 DNH 2 KFE HEK HEK HK Hk HEK HE HE HEK HEK HEK
3. HEK Hk Hk Hk HE HE HE HE HEK HEK HEK Hk Hk Hk HE HE HE HE HE HEK Hk Hk HE HE ERROR DETECTED POINT GROUP CHANGE DURING OPTIMISATION KFE HEK HEK Hk Hk HE HE HE HEK HEK HEK HEK Hk Hk Hk HE HE HE Hk HEK HEK HK Hk Hk Hk HE H HE HE HE HEK Hk Hk HE HE In such cases the optimisation may be completed by changing to the OPTXYZ algorithm using the RESTART option to retrieve the partially converged geometry from the Dumpfile 4 TRANSITION STATE AND GEOMETRY OPTIMISATION 7 as the starting point Note that OPTXYZ will not suffer from either ZMATRIX or Point group problems RESTART TITLE CAH2 3 21G ZMAT ANGS CA X 1 1 0 H 1 CAH 2 XCAH H 1 CAH 2 XCAH 3 180 0 VARIABLES CAH 2 148 HESSIAN 0 7 XCAH 75 0 HESSIAN 0 1 END RUNTYPE OPTXY VECTORS 1 ENTER 1 2 Particular care should be taken when dealing with the ZMATRIX specification for cyclic systems If this is specified such that a multiple bond is not explicitly defined then the initial step taken in an optimisation with either a poor starting geometry or ill defined Hessian can lead to unreasonable values for such a bond In the worst case the bond can become so compressed with a disastrous effect on the total energy that linear dependence will be correctly diagnosed Certainly in such cases the starting hessian should be taken from a smaller basis calculation STO 3G will do 3 PARAMETER ERRORS Possible errors may stem from the user not appreciating the workings of the optimisation modules
4. is shown below TRANSFORMED INTEGRAL PRE SORT CALLED AT 380 21 SECS TRANSFORMED 1 ELECTRON INTEGRALS RESTORED FROM SECTION 466 OF DUMPFILE STARTING AT BLOCK 1 OF ED3 x k SYMMETRY CHECK LARGEST FORBIDDEN H INTEGRAL 39 2 0 40457E 12 If there are inconsistencies between the ACTIVE and CONF specification this will almost cer tainly lead to a large value in the symmetry check in which case the diagonalisation results are probably meaningless 6 Table CI Calculations In previous releases of the code failing to allocate or pre compute the Table Cl data base lead to an operand range or related error at the outset of selection mode processing This situation has now been addressed with the program checking for correct assignment of the data base in the absence of the table directive 7 KNOWN BUGS ISSUES 10 7 Known Bugs Issues The list of know bugs as well as a list of feature requests is maintained on the GAMESS UK CCPForge site at http ccpforge cse rl ac uk tracker group_id 14 If you think you have encountered a bug that is not listed there please send an email to the gamess uk users list at http ccpforge cse rl ac uk mailman listinfo gamess uk users
5. CONTENTS i Computing for Science CFS Ltd CCLRC Daresbury Laboratory Generalised Atomic and Molecular Electronic Structure System Can ea I u l GAMESS UK USER S GUIDE and REFERENCE MANUAL Version 8 0 June 2008 PART 9 TROUBLE SHOOTING M F Guest J Kendrick J H van Lenthe and P Sherwood Copyright c 1993 2008 Computing for Science Ltd This document may be freely reproduced provided that it is reproduced unaltered and in its entirety Contents 1 Introduction 1 2 SCF Calculations 1 3 The role of the VECTORS and ENTER Directives 2 4 Transition State and Geometry Optimisation 5 4 1 Synchronous Transit Algorithm aoaaa a 5 Direct Cl Calculations 9 6 Table Cl Calculations 9 CONTENTS 7 Known Bugs Issues 10 1 INTRODUCTION 1 1 Introduction In this chapter we try to anticipate areas where the user may run into problems either from lack of appreciation of the workings of the program or from the author s failure to present a clear picture of requirements The items below are presented in no real order but reflect some of the known aspects of running the program that have given problems historically Again users are welcome to use the EMAIL addresses given in Chapter 1 in trying to resolve matters not discussed herein Note that the original use of error numbers has been replaced with textual messages designed to provide some insight into the problem causing the abort If the user meets an error conditi
6. as loaded under RUNTYPE OPTIMIZE or RUNTYPE SADDLE Both algorithms are based on the notion of a history file with any restart jobs working through the previous points along the optimisation pathway prior to contin uing execution at the interrupted point Anything which perturbs this process causing the history file as present on the Dumpfile to appear at odds with that generated from the data input will lead to a PARAMETER ERROR and termination of execution Ba sically once the optimisation has been initiated the user should not modify the input data other than through the introduction of the RESTART directive modification of the VECTORS specification etc Specifically modifying the ZMATRIX data or optimisation controls XTOL etc will lead to this error condition Let us consider one example where this might occur In the data files below we are using the results of a 3 21G optimisation as a basis for a larger TZVP optimisation The 3 21G data might appear as follows TITLE H20 3 21G ZMAT ANGSTROM 0 H 1 OH H 1 OH 2 HOH VARIABLES OH 0 956 HESSIAN 1 1 HOH 109 05 HESSIAN 0 2 END RUNTYPE OPTIMIZE VECTORS EXTGUESS ENTER 1 with the TZVP startup job below using the vectors and hessian from the split valence calculation 4 TRANSITION STATE AND GEOMETRY OPTIMISATION DUMPFILE ED3 300 TITLE H20 TZVP RESTARTING FROM 3 21G ZMAT ANGSTROM 0 H 1 OH H 1 OH 2 HOH VARIABLES OH 0 956 HESSIAN 1 1
7. ep 1 This is just the data specified in the startup job above TITLE WATER ZMAT ANGS 0 HiR HiR2A VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE OPTIMIZE ENTER 1 Data for Step 2 This now specifies the explicit section for the VECTORS the absence of which caused the problem in the restart job above 4 TRANSITION STATE AND GEOMETRY OPTIMISATION 5 RESTART TITLE WATER ZMAT ANGS 0 HiR HiR2A VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE CI DIRECT VECTORS 1 ENTER 1 4 Transition State and Geometry Optimisation 1 ZMATRIX and Point Group Problems One of the constraints in the z matrix formalism concerns potential problems in dealing with directly bonded angles of 180 The use of dummy X centres in circumventing these problems has been discussed in Part 3 88 3 Consider below the data file for optimisation of the CaH species where we are starting the optimisation with a bond angle of 150 TITLE CAH2 3 21G ZMAT ANGS CA H 1 CAH H 1 CAH 2 HCAH VARIABLES CAH 2 148 HESS 0 7 HCAH 150 0 HESS 0 1 END RUNTYPE OPTIMIZE VECTORS EXTGUESS ENTER 1 This job will fail with the diagnostics shown below taken from the failing output Z MATRIX ANGSTROMS AND DEGREES CD CENT ATOM N1 LENGTH N2 ALPHA N3 BETA 1 1 CA 2 4H 1 2 148195 1 3 3 H 1 2 148195 2 2 181 008 3 VARIABLE VALUE HESSIAN CAH
8. on characterised only by a number in the log please inform the author 2 SCF Calculations 1 The SCF aborts at the outset of processing this will almost certainly have arisen from supplying a two electron file that is not in the format expected Care must be taken in understanding the workings of the SUPER directive and the use of BYPASS 2 The SCF fails to converge with violent oscillations in the total energy increase the LEVEL parameters which should remove the problem 3 Large negative values for the energy the program assumes in constructing the Fock matrix that the density matrix transforms as the totally symmetric representation i e the wave function belongs to a non degenerate representation of the molecular point group If this is not the case the energy may collapse to a meaningless value If this occurs the user should modify the TAGs in the z matrix to lower the effective molecular point group 4 The SCF output reveals that the DIIS procedure has been initiated but the energy still fluctuates with DIIS appearing to push the energy up this may be indicative of trying to converge on an excited state particularly in closed shell calculations If the TESTER is small and the problem persists it may be advantageous to suppress DIIS through the data line DIIS OFF 5 The two electron integral file is lost or corrupted say in the middle of an optimisation use the REGEN parameter on the RESTART directive 6 The error me
9. rning to a previous point on the surface or 5 DIRECT CI CALCULATIONS 9 by exceptionally large values in the hessian matrix To avoid this reduce STEPMAX and if dealing with a closed shell system use LOCK 4 1 Synchronous Transit Algorithm 1 Too many z subspace iterations If many more than n 1 iterations are performed in any one z subspace minimisation then the parameter TOLMAX should be reduced 2 The optimisation will not converge e Check that the minima have been input correctly e Check that the starting guess is reasonable e Try calculating the full hessian ie specify TYPE 3 for all the variables it should have only one negative eigenvalue if the starting point is reasonable e Sometimes the linear search along the polynomial for a maximum causes problems which result in very large rises in energy Occasionally this may be overcome by increasing the VALUE parameter so that the linear search is not as rigorous check also that STEPMAX is not too large 5 Direct CI Calculations The internal symmetry manipulations within the Direct Cl module are largely hidden from the user although the re ordering performed to align both internal and external MOs into groups of common IRrep is recorded on the output It is possible for incorrect CONF specification to lead to problems and one of the best ways to monitor such effects is to look at the output from the pre sorting of the transformed integrals a typical example of which
10. section for use of VECTORS in the second step If no VECTORS directive is specified its default choice is a function of the perceived status of the processing e in startup job with no vectors available it will assume VECTORS ATOMS and proceed from an atomic startup e in a restart job it will assume that vectors are available and that such vectors are to be restored from a section on the Dumpfile If no such section is specified it will try to input those from a default section 410 This will lead to the diagnostic above Solution 1 Do not specify section numbers and let the programme defaults apply this is somewhat limiting since direct data driven access to the default section is not permitted by the user i e section 410 cannot be explicitly specified on a directive However in the present example the following data will suffice Data for Step 1 TITLE WATER ZMAT ANGS 0 HiR HiR2A 3 THE ROLE OF THE VECTORS AND ENTER DIRECTIVES 4 VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE OPTIMIZE ENTER Data for Step 2 RESTART TITLE WATER ZMAT ANGS 0 HiR HiR2A VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE CI DIRECT ENTER Solution 2 Use section number for explicit control of vector placing This will require use of the VECTORS directive in the second step to nominate section 1 Data for St
11. ssage ATTEMPTING TO RETRIEVE UNDEFINED DUMPFILE SECTION appears during vector generation in a restart job this may have arisen from specifying an incorrect section on the VECTORS directive but it is also indicative of the user having inadvertently omitted the RESTART directive or incorrectly allocated the Dumpfile to the Job 7 Confusion over Multiple Section Specification When performing open shell or GVB cal culations the user should be aware that the final list of output vectors refers to the canonicalised set i e to the set of MOs residing in the second section specified on the ENTER directive Confusion may arise when restoring MOs from the first section and assigning for example orbital symmetries from the SCF output 3 THE ROLE OF THE VECTORS AND ENTER DIRECTIVES 2 8 Symmetry contaminated molecular orbitals the SCF modules are fairly demanding in their attempts to maintain the symmetry classification of the MOs If the orbitals appear contaminated the cause may lie in the choice of an inappropriate starting set used to initialise the SCF 3 The role of the VECTORS and ENTER Directives GAMESS is reporting an error when trying to restore vectors in a restart job after a startup step has completed successfully referencing a section number unknown to the User Consider the following data that aims to first conduct an initial geometry optimisation followed in a subsequent step by a direct Cl calculation at the optimised geometr
12. y TITLE WATER ZMAT ANGS 0 HiR HiR2A VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE OPTIMIZE ENTER 1 with the following data specified in the restart job RESTART TITLE WATER ZMAT ANGS 0 HiR HiR2A VARIABLES R 0 9429786 HESSIAN 0 749526 A 105 9806454 HESSIAN 0 178413 END BASIS 6 31G RUNTYPE CI DIRECT ENTER 1 This restart will fail with the diagnostics shown below taken from the failing output MOLECULAR ORBITALS RESTORED FROM DUMPFILE 3 THE ROLE OF THE VECTORS AND ENTER DIRECTIVES 3 A VECTORS RESTORED FROM SECTION 410 OF DUMPFILE ON ED3 FO k k k kkk k k PROBLEM WITH SECTION aaoo k k k kkk k MPOS MTYPE BLOCK 410 3k k CURRENT INPUT LINE IS ENTER 1 HARAHAHAHAHARAHAEARAHARAHAEARAHAHARAEABAHABARABARAHAEARABABARABABABABAR ERROR DETECTED FO k k k kk ATTEMPTING TO RETRIEVE UNDEFINED DUMPFILE SECTION KEAEAEAHAHAHAHAHRAHAHARABARAHABARAHARAHAEARAHAHARAEABAHABABABARABAE AB K GAMESS UK ERROR ATTEMPTING TO RETRIEVE UNDEFINED DUMPFILE SECTION Diagnosis In the absence of section specification on the VECTORS and ENTER directive GAMESS will use a default section 410 for keeping the current set of vectors If the user specifies a section on the ENTER directive of the startup job section 1 above then that section MUST be specified in subsequent jobs for GAMESS will not redefine the initial ENTER section in the first step as being the
Download Pdf Manuals
Related Search