CCL:G: freezen dihedrals in five-membered rings

Hello Reynier,

If you have experimental values for the five dihedral angles, then these values come with error bars.  You cannot expect the five numbers to be compatible exactly as given.  Going back to the planar example, if your experimental error is +/- 1 degree, which would be very good accuracy for such a measurement, your experimental set could be {0,0,0,0,1}, which as we have seen, would be "impossible" when combined.

It sounds to me like what you would want to to is the following.  Fixing the known lengths of each bond in your ring and the 3-atom angles,  do not seek a geometry that matches all 5 dihedrals perfectly, but instead seek the geometry that minimizes the sum of the squares of the differences between its dihedrals and the target dihedrals.  This is a strictly mathematical problem that would not use Gaussian at all.  Once you have the "optimum" geometry, i.e. the geometry that most closely matches your five numbers according to a least-squares criterion, you can then do a single-point calc using Gaussian to see what its energy is.  Or, use the dihedrals that describe the least-squares geometry and do a constrained opt from there, but I don't think the geometry will change very much.

---- Original message ----

Date: Sat, 25 Sep 2010 15:11:33 +0200
From: "Reynier Suardiaz del Río reynier.suardiaz^^" <owner-chemistry-,>
Subject: CCL:G: freezen dihedrals in five-membered rings
To: "Anick, David " <david.anick-,>

Dear David

Thanks for your answer. You are right at all, the dihedrals are coupled and the value of one of them depends of the value os the others, so you can not aspire to have any five values of dihedral. Certainly, if you have a planar ring you can not change one dihedral (from its 0 degree value) and keeping the rest in that planar form because they have to change to be a possible geometry. What I am doing is giving to the dihedral experimental values, so this combinations of dihedral are possible. I am not changing one dihedral value and keeping the rest in their previous values (sorry if I was not clear enough). But, even changing the five dihedrals to a very similar possible values, the calculation is ending with that error.
I think gaussian is not recognizing this new combinations of dihedrals as possible geometries. If I keep frozen only two dihedral the calculation ends ok but the final obtained conformation have not the five dihedral values that I want (only the two frozen, the other three change a little). What I want to do is to obtain geometrically optimized conformations with the experimental values of the five dihedrals and them, calculate properties. In this way I can see how this property depends on the puckering of the ring.
I keep working on it, many thanks for your comments, they are very usefull to me.

All the best


On Sat, Sep 25, 2010 at 5:01 AM, <david.anick{}> wrote:
Dear Reynier,

What is happening is that when you have a ring, you can't just make the
dihedral angles anything you want, because there are mathematical relations among them.  To see this, consider a special case where all the atoms lie in a plane.  All five dihedral angles equal 0.  If you constrain all five angles to be zero, Gaussian is happy to optimize this.  Now suppose you change one of the angles, even if you only change it to +1 degree.  It becomes geometrically IMPOSSIBLE.  Because four of the five dihedrals are still zero, all five atoms are still forced to lie in a single plane.  Then a mathematical consequence is that the fifth dihedral is automatically zero also.  If you tell it to make that dihedral equal to +1, you are asking Gaussian to find a geometry that mathematically cannot exist.  Guess what: Gaussian cannot do it, and gives you an error message.  The error message is appropriate: "error imposing constraints".

If you want to change one dihedral angle a little, you must allow some of the other four dihedral angles to adapt to the change.  If you work a little with this, you will see that two dihedral angles essentially fix a five-member ring. (This is technically true only if the bond lengths and bond angles are also fixed, but it is hard to adapt bond lengths and angles to accommodate changing dihedral angles.)  Try this: change the dihedral angle you want to change, and remove three of the other constraints, so that you are specifying only two of the five dihedral angles.  I think Gaussian will be happy with that and will be able to converge.  If that works maybe you can try specifying three of the dihedrals, or conceivably, four, if the changes are very small.

You need to think about what is the question you are trying to answer.  If it's about the flexibility of the ring, your best approach may be to constrain just one dihedral, and let the rest of the ring adapt as it needs to.

I hope this has been helpful.
David Anick PhD MD

---- Original message ----

Date: Fri, 24 Sep 2010 21:27:14 +0200
From: "Reynier Suardiaz del Río reynier.suardiaz!^!" <owner-chemistry{}>

Subject: CCL:G: freezen dihedrals in five-membered rings
To: "Anick, David " <david.anick{}>

Dear D. Close

Many thanks for your answer, you was right. I typed the values of the dihedrals with 6 decimal places exactly matching with those of the input structure. Doing this, the geometry optimization have finished without problem in a few iterations. Now what I would like to do is the following: I want to generate diferent conformations of this furanose ring by changing the dihedrals (between permitted values without breaking of the ring) and partially optimize this structures (obtained by slightly changing the dihedral values) and keeping frozen the five dihedrals. When I try to do this using redundant coordinates in gaussian I obtained the same error message than before:

 Iteration 99 RMS(Cart)=  0.00005822 RMS(Int)=  0.00955580
 Iteration100 RMS(Cart)=  0.00005748 RMS(Int)=  0.00959825
 New curvilinear step not converged.
 Error imposing constraints
 Error termination via Lnk1e in C:\G03W\l101.exe at Fri Sep 24 21:05:06 2010.
 Job cpu time:  0 days  0 hours  0 minutes  1.0 seconds.
 File lengths (MBytes):  RWF=      7 Int=      0 D2E=      0 Chk=      8 Scr=      1

even if I only change one dihedral from its original value (at the input geometry) in less than one degree,  the calculation ends with the above error message.
Does anybody knows how to do this in gaussian, I mean, changing the dihedral angles of a five memebered ring (from its text input file) and to performe a partial geometrical optimization with diferent dihedral angles, other than the one of the input geometry?

any comment or suggestions are welcome.

thanks in advance and with very best regards


2010/9/22 Close, David M. CLOSED~! <owner-chemistry]*[>

Sent to CCL by: "Close, David M." [CLOSED#,]
 There is no limit to how many dihedrals you can freeze.  The problem is that you typed something wrong.  Notice that the program tried 99 iterations to fit you frozen coordiate information into the optimization routine.

Either you connected the coordinates incorrectly, or did not have enough precision in the frozen coordinate.
So if the input line has something like 10 5 6 8   31.3, first look at the string 10 5 6 8 and make sure this is correct.  The use a graphics program to examine the actual dihedral geometry.  Run through the 4 atoms in the string 10 5 6 8 and see what the graphic program thinks the dihedral angle actually is.  Copy the value to 5-6 decimal places and re enter the data.

 If this doesn't work, then you have to use trial and error.  You said that freezing 2 dihedrals works.  But how many iterations did it take?  I would expect only 2-3.  If more, then refine the coordinates, and then add a third frozen dihedral.  You can quickly find the offending entry when the optimization routine bombs.

 Regards, Dave Close.

> From:[A] [[A]] on behalf of Reynier Suard az reynier.suardiaz(a) [owner-chemistry[A]]

Sent: Wednesday, September 22, 2010 10:49 AM
To: Close, David M.
Subject: CCL:G: freezen dihedrals in five-membered rings

Sent to CCL by: "Reynier  Suard  az" [reynier.suardiaz]_[]
Dear All

I want to generate a lot of conformations of furanose ring (or cyclopentane?) and later partially optimize them but keeping frozen the dihedral angles. I am trying to use redundant coordinates in gaussian writing at the end of the input gaussian file the desired dihedrals to keep frozen. I am receiving an error message when i try to keep frozen more than two dihedral angles (of the ring) at the same time. For example if I try to froze the five dihedrals of the ring I get the following message:

 Iteration 99 RMS(Cart)=  0.00001156 RMS(Int)=  0.00309967
 Iteration100 RMS(Cart)=  0.00001134 RMS(Int)=  0.00310385
 New curvilinear step not converged.
 Error imposing constraints
 Error termination via Lnk1e in C:\G03W\l101.exe at Mon Sep 20 17:48:17 2010.
 Job cpu time:  0 days  0 hours  0 minutes  1.0 seconds.
 File lengths (MBytes):  RWF=      7 Int=      0 D2E=      0 Chk=      7 Scr=      1

I receive this error message even when I try to freeze the dihedral at the same value they already have in the initial structure.

Is not possible what am I trying to do? How can I overcome this problem with gaussian? Is there any other possibility to do this kind of partial optimization in five-membered rings? Note that I can not freeze all the dihedrals using optimization in internal coordinates (opt=z-matrix with a separate input section of "constants") because of to define a z-matrix of a five-membered ring are necesary only two dihedrals, so I have to use redundants.

any comments or sugestions would be appreciatte

thanks in advance and very best regards