Molcas Forum

Addiw7

Member

Registered: 2016-05-24

Posts: 11

QM/MM calculations with Molcas8.0-UU and Tinker

Dear Molcas Experts,

I have succesfully installed Molcas8.0-UU (15.06.18_CentOS_6.6_x86_64) and installed and patched Tinker 6.3.3
with
./molcas get_tinker

I run calculations and I get an error that you may see in output file attached. I also attach the input files.

http://www.speedyshare.com/kyw6r/input- … les.tar.gz

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: QM/MM calculations with Molcas8.0-UU and Tinker

I don't know if this will help, but try with shorter path names. Some internal strings have some length limits and you may be breaking them...

Addiw7

Member

Registered: 2016-05-24

Posts: 11

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Thank you for responding, but it does not do any good.
However, I guess I am getting closer to the solution, because

*.Tinker.log
file says:
This is a QM/MM calculation
  QMMMSETUP -- Wrong QMMM-ELECTROSTATICS

TINKER is Unable to Continue; Terminating the Current Calculation

Thus, my question now is whether you know that there are some deficiencies and/or restrictions for using precompiled Molcas8.0-UU with Tinker for QM/MM calculations?

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Not that I know, at least not more than with the "normal" Molcas 8.0. That looks like a Tinker error.

niko

Member

From: Marseille

Registered: 2015-11-08

Posts: 59

Website

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Hi,
Could you just remove the "QMMM-ELECTROSTATICS ESPF" line in the key file?

Addiw7

Member

Registered: 2016-05-24

Posts: 11

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Hi,

You were right niko! I got those inputs from a friend of mine and I guess that they worked fine when I used them for calculations with licensed Molcas.
Now, we will check whether results for the same system and calculations differ or not.
Thanks anyway!

Addiw7

Member

Registered: 2016-05-24

Posts: 11

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Hi,

So, the results that I get now are completely wrong, because the electrostatic interactions are not calculated properly.

I wrote in my first message that I used patch for Tinker 6.3.3. In this patch file, there are lines saying:

+c     Select the QM/MM electrostatics interaction mode
+c
+         else if (keyword(1:20) .eq. 'QMMM-ELECTROSTATICS ') then
+            call gettext (record,keyword,next)
+            call upcase (keyword)
+            if (keyword(1:5) .eq. 'NONE ') then
+               e4qmmm = -1
+            else
+               write (iout,*) ' QMMMSETUP -- Wrong QMMM-ELECTROSTATICS'
+               call fatal
+            end if

which clearly means that if I use ESPF I get an error message and in fact I get it in my *.Tinker.log file.
So how can I overcome it?
In the same file, it is said:

+c     e4qmmm         count or not some QM/MM electrostatic interactions
+c                     -1  ->  none, aka mechanical embedding
+c                      0  ->  none but will be computed elsewhere
+c                      3  ->  includes the QM/MM ones (microiteration or MD cases)

so how can I activate this e4qmmm=3 ?
I cannot figure that out from what the file says.

niko

Member

From: Marseille

Registered: 2015-11-08

Posts: 59

Website

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Hi,
The ESPF scheme is handled automatically, this is why I said in my previous post that you need to erase the QMMM-ELECTROSTATICS keyword in the Tinker key file.

e4qmmm needs to be 0 when QM/MM electrostatics interactions are computed on the QM side (ie Molcas) and 3 when they are computed on the MM side (Tinker), the latter occurring for instance during MM microiterations when QM atoms are electrostatically represented by ESPF multipoles. But again, the e4qmmm value is set automatically.

Could you explain why you write "results are completely wrong"? Examples are welcome!

Addiw7

Member

Registered: 2016-05-24

Posts: 11

Re: QM/MM calculations with Molcas8.0-UU and Tinker

Hi,

I am sorry for "results are completely wrong". I wrote so because the results of ESPF calculations that I got using MolcasUU-8.0 are completely different from those that
we obtained previously with Molcas 7.9 that we compiled ourselves but now our license expired.

However, I talked to my supervisor and he said that it does not mean that the results are wrong. Perhaps Molcas 8.0 handles the electrostatics calculations in different
manner than Molcas 7.9 does. So we will run more tests.

Again, sorry for confusing and not well-thought posts. I will let you know when I find something new.

niko

Member

From: Marseille

Registered: 2015-11-08

Posts: 59

Website

Re: QM/MM calculations with Molcas8.0-UU and Tinker

For the record, since Molcas 8 all the contributions to the QM/MM electrostatics are handled by the ESPF scheme while it was not the case in previous versions (direct interactions between QM and the MM atoms defined in &Gateway). Now, the results should not be so different ...