Re: Structure program

Posted by Evgeniy on May 11, 2009 at 16:22:03:

In Reply to: Structure program posted by Oscar Rubio POns on April 21, 2009 at 16:21:11:

: Dear MOLCAS USERS

: I have found something strange in the optimization procedure. I am using MOLCAS 7.2 and running a CASSCF (14/12) ground state optimization.

: I have done a long serie of calcualtion and after a while I am not very confident about the optimization
: It seems a some step MOLCAS is reading the "old" data and mixing steps. I have done 7 optimizations.
: After the 7th, the calculation stops, with no reason, and I have checked the partial ouputs, I have the surprise that the step 7, is with a past (less optimized) geometry. Can anyone explain me what it is wrong ?

: The partial outputs, please take a look to the output from Step 7.

: -rw-r--r-- 1 1319453 2009-04-18 03:03 Structure.1.output
: -rw-r--r-- 1 120958 2009-04-18 15:56 Structure.2.output
: -rw-r--r-- 1 101366 2009-04-19 07:21 Structure.3.output
: -rw-r--r-- 1 101988 2009-04-19 21:07 Structure.4.output
: -rw-r--r-- 1 102636 2009-04-20 15:40 Structure.5.output
: -rw-r--r-- 1 103392 2009-04-21 03:20 Structure.6.output
: -rw-r--r-- 1 1 90002 2009-04-21 15:40 Structure.7.output

:
: Here I paste the last output lines....

: --Step 5----

: 60 -1662.30749592 -0.00002660 0.007705-0.002457 nrc060 -0.010407 nrc065 -1662.30752390 RS-RFO BFGS 0
: 61 -1662.30752637 -0.00003045 0.009196 0.002798 nrc062 -0.002304 nrc064 -1662.30753764 RS-RFO BFGS 0
: 62 -1662.30754721 -0.00002084 0.009228-0.002985 nrc031 0.014509 nrc009 -1662.30759726 RS-RFO BFGS 0
: 63 -1662.30760443 -0.00005722 0.007771 0.002721 nrc062 -0.009047 nrc059 -1662.30763546 RS-RFO BFGS 0
: 64 -1662.30764848 -0.00004405 0.008989 0.002799 nrc062 -0.012614 nrc076 -1662.30768194 RS-RFO BFGS 0
: 65 -1662.30770034 -0.00005187 0.009515 0.002843 nrc062 -0.032536 nrc076 -1662.30777060 RS-RFO BFGS 0
: 66 -1662.30777128 -0.00007093 0.009572 0.003062 nrc059 0.003805 nrc059 -1662.30779674 RS-RFO BFGS 0
: 67 -1662.30781660 -0.00004532 0.008053 0.002821 nrc059 0.016142 nrc062 -1662.30787575 RS-RFO BFGS 0
: 68 -1662.30788520 -0.00006860 0.007521-0.002222 nrc051 -0.016572 nrc065 -1662.30792743 RS-RFO BFGS 0
: 69 -1662.30793918 -0.00005398 0.010002-0.002996 nrc051 -0.019439 nrc076 -1662.30797026 RS-RFO BFGS 0
: 70 -1662.30797597 -0.00003679 0.011466-0.003677 nrc051 -0.017248 nrc076 -1662.30801173 RS-RFO BFGS 0
: 71 -1662.30802984 -0.00005387 0.010574-0.004227 nrc067 -0.032305 nrc076 -1662.30811525 RS-RFO BFGS 0
: 72 -1662.30813949 -0.00010964 0.008060-0.003533 nrc050 -0.005018 nrc076 -1662.30816640 RS-RFO BFGS 0

: +----------------------------------+----------------------------------+
: + Cartesian Displacements + Gradient in internals +
: + Value Threshold Converged? + Value Threshold Converged? +
: +-----+----------------------------------+----------------------------------+
: + RMS + 0.4188E-02 0.1200E-02 No + 0.9126E-03 0.3000E-03 No +
: +-----+----------------------------------+----------------------------------+
: + Max + 0.7494E-02 0.1800E-02 No + 0.3533E-02 0.4500E-03 No +
: +-----+----------------------------------+----------------------------------+

: Convergence not reached yet!
: -----Step 6-----

: +----------------------------------+----------------------------------+
: + Cartesian Displacements + Gradient in internals +
: + Value Threshold Converged? + Value Threshold Converged? +
: +-----+----------------------------------+----------------------------------+
: + RMS + 0.7721E-02 0.1200E-02 No + 0.9290E-03 0.3000E-03 No +
: +-----+----------------------------------+----------------------------------+
: + Max + 0.1284E-01 0.1800E-02 No + 0.3013E-02 0.4500E-03 No +
: +-----+----------------------------------+----------------------------------+

: Convergence not reached yet!
: -- Step 7---

: 56 -1662.30731809 -0.00009414 0.009414-0.003663 nrc071 -0.045143 nrc074 -1662.30744699 RS-RFO BFGS 0
: 57 -1662.30734762 -0.00002953 0.014887-0.003514 nrc059 0.016175 nrc074 -1662.30737572 RS-RFO BFGS 0
: 58 -1662.30742840 -0.00008078 0.010655-0.002741 nrc059 -0.004164 nrc059 -1662.30745851 RS-RFO BFGS 0
: 59 -1662.30746932 -0.00004092 0.008085-0.002457 nrc060 -0.006704 nrc074 -1662.30748826 RS-RFO BFGS 0
: 60 -1662.30749592 -0.00002660 0.007705-0.002457 nrc060 -0.010407 nrc065 -1662.30752390 RS-RFO BFGS 0
: 61 -1662.30752637 -0.00003045 0.009196 0.002798 nrc062 -0.002304 nrc064 -1662.30753764 RS-RFO BFGS 0
: 62 -1662.30754721 -0.00002084 0.009228-0.002985 nrc031 0.014509 nrc009 -1662.30759726 RS-RFO BFGS 0

: +----------------------------------+----------------------------------+
: + Cartesian Displacements + Gradient in internals +
: + Value Threshold Converged? + Value Threshold Converged? +
: +-----+----------------------------------+----------------------------------+
: + RMS + 0.9321E-02 0.1200E-02 No + 0.1045E-02 0.3000E-03 No +
: +-----+----------------------------------+----------------------------------+
: + Max + 0.1431E-01 0.1800E-02 No + 0.2985E-02 0.4500E-03 No +
: +-----+----------------------------------+----------------------------------+

: Convergence not reached yet!

:
: I do not know why the last step (Step 7) is the last one, with a higher energy, and worst thresholds...
: Any explanation ?
: Which geometry is using MOLCAS to re-start the calculation ?
: Is this a bug?

: Many thanks
: Oscar Rubio

Dear Oscar,

I have encountered such kind of a problem several times with Molcas 7.2 (see here:
/molcas/wwwboard/messages/1068.html). I have had a suspicion that it is related to some memory leaks although Valera Veryazov said it was hardly possible. Anyway the problem has remained. You can circumvent it if you run the modules in the optimization loop one by one untill it is converged. Of course it is better to use a script which would launch the modules sequentially and would check whether the convergence is reached. On the other hand, fom your input I noticed that the optimization started diverging/oscillating after step 68. I guess your system is quite large. In this case you might not get it converge with the RFO scheme. I don't know why but it breaks down when it approaches the optimal geometry in the case of large systems. I would very appreciate if the developers could check this out!

Follow Ups:

Post a Followup

Name:
E-Mail:

Subject:

if B is 1s22s22p1, what is Li?

Passfield:

Comments:
: : Dear MOLCAS USERS : : I have found something strange in the optimization procedure. I am using MOLCAS 7.2 and running a CASSCF (14/12) ground state optimization. : : I have done a long serie of calcualtion and after a while I am not very confident about the optimization : : It seems a some step MOLCAS is reading the "old" data and mixing steps. I have done 7 optimizations. : : After the 7th, the calculation stops, with no reason, and I have checked the partial ouputs, I have the surprise that the step 7, is with a past (less optimized) geometry. Can anyone explain me what it is wrong ? : : The partial outputs, please take a look to the output from Step 7. : : -rw-r--r-- 1 1319453 2009-04-18 03:03 Structure.1.output : : -rw-r--r-- 1 120958 2009-04-18 15:56 Structure.2.output : : -rw-r--r-- 1 101366 2009-04-19 07:21 Structure.3.output : : -rw-r--r-- 1 101988 2009-04-19 21:07 Structure.4.output : : -rw-r--r-- 1 102636 2009-04-20 15:40 Structure.5.output : : -rw-r--r-- 1 103392 2009-04-21 03:20 Structure.6.output : : -rw-r--r-- 1 1 90002 2009-04-21 15:40 Structure.7.output : : : : Here I paste the last output lines.... : : --Step 5---- : : 60 -1662.30749592 -0.00002660 0.007705-0.002457 nrc060 -0.010407 nrc065 -1662.30752390 RS-RFO BFGS 0 : : 61 -1662.30752637 -0.00003045 0.009196 0.002798 nrc062 -0.002304 nrc064 -1662.30753764 RS-RFO BFGS 0 : : 62 -1662.30754721 -0.00002084 0.009228-0.002985 nrc031 0.014509 nrc009 -1662.30759726 RS-RFO BFGS 0 : : 63 -1662.30760443 -0.00005722 0.007771 0.002721 nrc062 -0.009047 nrc059 -1662.30763546 RS-RFO BFGS 0 : : 64 -1662.30764848 -0.00004405 0.008989 0.002799 nrc062 -0.012614 nrc076 -1662.30768194 RS-RFO BFGS 0 : : 65 -1662.30770034 -0.00005187 0.009515 0.002843 nrc062 -0.032536 nrc076 -1662.30777060 RS-RFO BFGS 0 : : 66 -1662.30777128 -0.00007093 0.009572 0.003062 nrc059 0.003805 nrc059 -1662.30779674 RS-RFO BFGS 0 : : 67 -1662.30781660 -0.00004532 0.008053 0.002821 nrc059 0.016142 nrc062 -1662.30787575 RS-RFO BFGS 0 : : 68 -1662.30788520 -0.00006860 0.007521-0.002222 nrc051 -0.016572 nrc065 -1662.30792743 RS-RFO BFGS 0 : : 69 -1662.30793918 -0.00005398 0.010002-0.002996 nrc051 -0.019439 nrc076 -1662.30797026 RS-RFO BFGS 0 : : 70 -1662.30797597 -0.00003679 0.011466-0.003677 nrc051 -0.017248 nrc076 -1662.30801173 RS-RFO BFGS 0 : : 71 -1662.30802984 -0.00005387 0.010574-0.004227 nrc067 -0.032305 nrc076 -1662.30811525 RS-RFO BFGS 0 : : 72 -1662.30813949 -0.00010964 0.008060-0.003533 nrc050 -0.005018 nrc076 -1662.30816640 RS-RFO BFGS 0 : : +----------------------------------+----------------------------------+ : : + Cartesian Displacements + Gradient in internals + : : + Value Threshold Converged? + Value Threshold Converged? + : : +-----+----------------------------------+----------------------------------+ : : + RMS + 0.4188E-02 0.1200E-02 No + 0.9126E-03 0.3000E-03 No + : : +-----+----------------------------------+----------------------------------+ : : + Max + 0.7494E-02 0.1800E-02 No + 0.3533E-02 0.4500E-03 No + : : +-----+----------------------------------+----------------------------------+ : : Convergence not reached yet! : : -----Step 6----- : : 60 -1662.30749592 -0.00002660 0.007705-0.002457 nrc060 -0.010407 nrc065 -1662.30752390 RS-RFO BFGS 0 : : 61 -1662.30752637 -0.00003045 0.009196 0.002798 nrc062 -0.002304 nrc064 -1662.30753764 RS-RFO BFGS 0 : : 62 -1662.30754721 -0.00002084 0.009228-0.002985 nrc031 0.014509 nrc009 -1662.30759726 RS-RFO BFGS 0 : : 63 -1662.30760443 -0.00005722 0.007771 0.002721 nrc062 -0.009047 nrc059 -1662.30763546 RS-RFO BFGS 0 : : 64 -1662.30764848 -0.00004405 0.008989 0.002799 nrc062 -0.012614 nrc076 -1662.30768194 RS-RFO BFGS 0 : : 65 -1662.30770034 -0.00005187 0.009515 0.002843 nrc062 -0.032536 nrc076 -1662.30777060 RS-RFO BFGS 0 : : 66 -1662.30777128 -0.00007093 0.009572 0.003062 nrc059 0.003805 nrc059 -1662.30779674 RS-RFO BFGS 0 : : 67 -1662.30781660 -0.00004532 0.008053 0.002821 nrc059 0.016142 nrc062 -1662.30787575 RS-RFO BFGS 0 : : 68 -1662.30788520 -0.00006860 0.007521-0.002222 nrc051 -0.016572 nrc065 -1662.30792743 RS-RFO BFGS 0 : : 69 -1662.30793918 -0.00005398 0.010002-0.002996 nrc051 -0.019439 nrc076 -1662.30797026 RS-RFO BFGS 0 : : 70 -1662.30797597 -0.00003679 0.011466-0.003677 nrc051 -0.017248 nrc076 -1662.30801173 RS-RFO BFGS 0 : : 71 -1662.30802984 -0.00005387 0.010574-0.004227 nrc067 -0.032305 nrc076 -1662.30811525 RS-RFO BFGS 0 : : 72 -1662.30813949 -0.00010964 0.008060-0.003533 nrc050 -0.005018 nrc076 -1662.30816640 RS-RFO BFGS 0 : : 73 -1662.30818124 -0.00004176 0.008152-0.003013 nrc050 -0.009029 nrc076 -1662.30821966 RS-RFO BFGS 0 : : +----------------------------------+----------------------------------+ : : + Cartesian Displacements + Gradient in internals + : : + Value Threshold Converged? + Value Threshold Converged? + : : +-----+----------------------------------+----------------------------------+ : : + RMS + 0.7721E-02 0.1200E-02 No + 0.9290E-03 0.3000E-03 No + : : +-----+----------------------------------+----------------------------------+ : : + Max + 0.1284E-01 0.1800E-02 No + 0.3013E-02 0.4500E-03 No + : : +-----+----------------------------------+----------------------------------+ : : Convergence not reached yet! : : -- Step 7--- : : 56 -1662.30731809 -0.00009414 0.009414-0.003663 nrc071 -0.045143 nrc074 -1662.30744699 RS-RFO BFGS 0 : : 57 -1662.30734762 -0.00002953 0.014887-0.003514 nrc059 0.016175 nrc074 -1662.30737572 RS-RFO BFGS 0 : : 58 -1662.30742840 -0.00008078 0.010655-0.002741 nrc059 -0.004164 nrc059 -1662.30745851 RS-RFO BFGS 0 : : 59 -1662.30746932 -0.00004092 0.008085-0.002457 nrc060 -0.006704 nrc074 -1662.30748826 RS-RFO BFGS 0 : : 60 -1662.30749592 -0.00002660 0.007705-0.002457 nrc060 -0.010407 nrc065 -1662.30752390 RS-RFO BFGS 0 : : 61 -1662.30752637 -0.00003045 0.009196 0.002798 nrc062 -0.002304 nrc064 -1662.30753764 RS-RFO BFGS 0 : : 62 -1662.30754721 -0.00002084 0.009228-0.002985 nrc031 0.014509 nrc009 -1662.30759726 RS-RFO BFGS 0 : : +----------------------------------+----------------------------------+ : : + Cartesian Displacements + Gradient in internals + : : + Value Threshold Converged? + Value Threshold Converged? + : : +-----+----------------------------------+----------------------------------+ : : + RMS + 0.9321E-02 0.1200E-02 No + 0.1045E-02 0.3000E-03 No + : : +-----+----------------------------------+----------------------------------+ : : + Max + 0.1431E-01 0.1800E-02 No + 0.2985E-02 0.4500E-03 No + : : +-----+----------------------------------+----------------------------------+ : : Convergence not reached yet! : : : : I do not know why the last step (Step 7) is the last one, with a higher energy, and worst thresholds... : : Any explanation ? : : Which geometry is using MOLCAS to re-start the calculation ? : : Is this a bug? : : Many thanks : : Oscar Rubio : : Dear Oscar, : I have encountered such kind of a problem several times with Molcas 7.2 (see here: : /molcas/wwwboard/messages/1068.html). I have had a suspicion that it is related to some memory leaks although Valera Veryazov said it was hardly possible. Anyway the problem has remained. You can circumvent it if you run the modules in the optimization loop one by one untill it is converged. Of course it is better to use a script which would launch the modules sequentially and would check whether the convergence is reached. On the other hand, fom your input I noticed that the optimization started diverging/oscillating after step 68. I guess your system is quite large. In this case you might not get it converge with the RFO scheme. I don't know why but it breaks down when it approaches the optimal geometry in the case of large systems. I would very appreciate if the developers could check this out!

[ Follow Ups ] [ Post Followup ] [ Molcas user's WWWBoard ]