[gmx-users] relative weight between the direct space sum and reciprocal space sum

[Argyrios Karatrantos]

HI,

according to the 3.2 gromacs manual , the ewald_rtol parameter is the relative strength of the electrostatic interaction at the cut off, decreasing this gives a more accurate direct sum, but a less accurate reciprocal sum(p. 81).
The beta parameter is the parameter that determines the relative weight between the direct sum and reciprocal sum(p.55) but i don't think you can set the beta parameter in gromacs
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20061206/edbde621/attachment.html>

[Mark Abraham]

Post by Argyrios Karatrantos
HI,
according to the 3.2 gromacs manual , the ewald_rtol parameter is the
relative strength of the electrostatic interaction at the cut off,
decreasing this gives a more accurate direct sum, but a less accurate
reciprocal sum(p. 81).
The beta parameter is the parameter that determines the relative weight
between the direct sum and reciprocal sum(p.55) but i don't think you
can set the beta parameter in gromacs

As I said yesterday, you can choose its value, but you do so indirectly
using rcoulomb and ewald_rtol. ewald_rtol controls the magnitude of the
error in the direct space summation according to

ewald_rtol >= erfc( beta * r_{i,j,n}) / r_{i,j,n}

for all r_{i,j,n} > r_coulomb

Here, beta is a free parameter, and is chosen to satisfy the equality of
the first equation. This value of beta is then used in the remaining
Ewald terms.

Further, gromacs reports the value of beta being used in the .log file
(search for beta)

Mark

[Janne Hirvi]

Hello!

Yes, I think you cant set the beta parameter directly but you can calculate it
from the equation

erfc(beta*r)/r = ewald_rtol

at the cut-off (=rcoulomb) and so set the ewald_rtol parameter to match to the
wanted beta parameter.

Janne

Post by Argyrios Karatrantos
HI,
according to the 3.2 gromacs manual , the ewald_rtol parameter is the
relative strength of the electrostatic interaction at the cut off, decreasing
this gives a more accurate direct sum, but a less accurate reciprocal sum(p.
81).
The beta parameter is the parameter that determines the relative weight
between the direct sum and reciprocal sum(p.55) but i don't think you can set
the beta parameter in gromacs

[Janne Hirvi]

Hello!

Just a simple question - Do you have [pairs]-section in your topology file?

Janne

Dear GMX-users,
Here are a followup and more details for my problem. I really hope to get
through this, so I can move on ...
I report the electrostatic energy (Uq, kJ/mol) for a molecule using 3
codes.
1. Gromacs 3.3
[default]
1 1 yes 0.5 FudgeQQ
FudgeQQ [moleculetype] Uq
0.0, 0.8333, or 1.0 mol 3 116.050
0.0, 0.8333, or 1.0 mol 2 -413.048
2. DLPOLY with scaling factor of 0.83333
Uq = -324.68
3. My own code
Scaling factor Uq
0.0 115.99
0.83333 -324.68
1.0 -412.81
Apparently, Gromacs gives the value without 1-4 interactions with "mol 3",
and the value for full 1-4 interactions with "mol 2", regardless of
fudgeQQ.
Did I not do something right when using Gromacs?
Thanks A LOT for your help! BTW: the Gromacs website seems down.
Lianqing

Dear GMX-users,
I'm doing normal mode analysis for one molecule and found the calculated
electrostatic energy (at t=0) doesn't change at all regardless of the
value for fudgeQQ. (I used "1 1 yes 0.5 0.83333" for [default] and
"molecule 2" for [moleculetype].) If I use "molecule 3", the value will
change but is still independent of fudgeQQ.
Any idea what the problem is? I'll be happy to send you the input files if
you need them.
Thanks a lot!
Lianqing

[Lianqing Zheng]

Thanks, Janne!

No, I tried not having [pairs] or having empty [pairs] because the manual
says it's not necessary if "generate pairs" = "yes".

Lianqing

Post by Janne Hirvi
Hello!
Just a simple question - Do you have [pairs]-section in your topology file?
Janne

Dear GMX-users,
Here are a followup and more details for my problem. I really hope to get
through this, so I can move on ...
I report the electrostatic energy (Uq, kJ/mol) for a molecule using 3
codes.
1. Gromacs 3.3
[default]
1 1 yes 0.5 FudgeQQ
FudgeQQ [moleculetype] Uq
0.0, 0.8333, or 1.0 mol 3 116.050
0.0, 0.8333, or 1.0 mol 2 -413.048
2. DLPOLY with scaling factor of 0.83333
Uq = -324.68
3. My own code
Scaling factor Uq
0.0 115.99
0.83333 -324.68
1.0 -412.81
Apparently, Gromacs gives the value without 1-4 interactions with "mol 3",
and the value for full 1-4 interactions with "mol 2", regardless of
fudgeQQ.
Did I not do something right when using Gromacs?
Thanks A LOT for your help! BTW: the Gromacs website seems down.
Lianqing

Dear GMX-users,
I'm doing normal mode analysis for one molecule and found the calculated
electrostatic energy (at t=0) doesn't change at all regardless of the
value for fudgeQQ. (I used "1 1 yes 0.5 0.83333" for [default] and
"molecule 2" for [moleculetype].) If I use "molecule 3", the value will
change but is still independent of fudgeQQ.
Any idea what the problem is? I'll be happy to send you the input files if
you need them.
Thanks a lot!
Lianqing

_______________________________________________
gmx-users mailing list gmx-users at gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-request at gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php

[Janne Hirvi]

Hi!

I am not exactly sure but I think that there should be [pairs] section with all
pairs named and the function number (=1) included but without any parameters
which are then generated by Gromacs if "generate pairs" = "yes". You can check
from the .log -file if you have any 1-4 Coulombic (or 1-4 Lennard-Jones)
interactions included in your case.

Janne

Post by Lianqing Zheng
Thanks, Janne!
No, I tried not having [pairs] or having empty [pairs] because the manual
says it's not necessary if "generate pairs" = "yes".
Lianqing

Post by Janne Hirvi
Hello!
Just a simple question - Do you have [pairs]-section in your topology file?
Janne

Dear GMX-users,
Here are a followup and more details for my problem. I really hope to get
through this, so I can move on ...
I report the electrostatic energy (Uq, kJ/mol) for a molecule using 3
codes.
1. Gromacs 3.3
[default]
1 1 yes 0.5 FudgeQQ
FudgeQQ [moleculetype] Uq
0.0, 0.8333, or 1.0 mol 3 116.050
0.0, 0.8333, or 1.0 mol 2 -413.048
2. DLPOLY with scaling factor of 0.83333
Uq = -324.68
3. My own code
Scaling factor Uq
0.0 115.99
0.83333 -324.68
1.0 -412.81
Apparently, Gromacs gives the value without 1-4 interactions with "mol

3",

Post by Janne Hirvi

and the value for full 1-4 interactions with "mol 2", regardless of
fudgeQQ.
Did I not do something right when using Gromacs?
Thanks A LOT for your help! BTW: the Gromacs website seems down.
Lianqing

Dear GMX-users,
I'm doing normal mode analysis for one molecule and found the calculated
electrostatic energy (at t=0) doesn't change at all regardless of the
value for fudgeQQ. (I used "1 1 yes 0.5 0.83333" for [default] and
"molecule 2" for [moleculetype].) If I use "molecule 3", the value will
change but is still independent of fudgeQQ.
Any idea what the problem is? I'll be happy to send you the input files

if

Post by Janne Hirvi

you need them.
Thanks a lot!
Lianqing

[Lianqing Zheng]

Thanks a lot, Janne! I finally got it! Thanks for your inspiration! I
added 1-4 pairs in [pairs] and then got the right values. I thought
Gromacs would automatically find 1-4 pairs.

Lianqing

Post by Janne Hirvi
Hi!
I am not exactly sure but I think that there should be [pairs] section with all
pairs named and the function number (=1) included but without any parameters
which are then generated by Gromacs if "generate pairs" = "yes". You can check
from the .log -file if you have any 1-4 Coulombic (or 1-4 Lennard-Jones)
interactions included in your case.
Janne

Post by Lianqing Zheng
Thanks, Janne!
No, I tried not having [pairs] or having empty [pairs] because the manual
says it's not necessary if "generate pairs" = "yes".
Lianqing

Post by Janne Hirvi
Hello!
Just a simple question - Do you have [pairs]-section in your topology file?
Janne

Dear GMX-users,
Here are a followup and more details for my problem. I really hope to get
through this, so I can move on ...
I report the electrostatic energy (Uq, kJ/mol) for a molecule using 3
codes.
1. Gromacs 3.3
[default]
1 1 yes 0.5 FudgeQQ
FudgeQQ [moleculetype] Uq
0.0, 0.8333, or 1.0 mol 3 116.050
0.0, 0.8333, or 1.0 mol 2 -413.048
2. DLPOLY with scaling factor of 0.83333
Uq = -324.68
3. My own code
Scaling factor Uq
0.0 115.99
0.83333 -324.68
1.0 -412.81
Apparently, Gromacs gives the value without 1-4 interactions with "mol

3",

Post by Janne Hirvi

and the value for full 1-4 interactions with "mol 2", regardless of
fudgeQQ.
Did I not do something right when using Gromacs?
Thanks A LOT for your help! BTW: the Gromacs website seems down.
Lianqing

Dear GMX-users,
I'm doing normal mode analysis for one molecule and found the calculated
electrostatic energy (at t=0) doesn't change at all regardless of the
value for fudgeQQ. (I used "1 1 yes 0.5 0.83333" for [default] and
"molecule 2" for [moleculetype].) If I use "molecule 3", the value will
change but is still independent of fudgeQQ.
Any idea what the problem is? I'll be happy to send you the input files

if

Post by Janne Hirvi

you need them.
Thanks a lot!
Lianqing

_______________________________________________
gmx-users mailing list gmx-users at gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-request at gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php

[chris.neale]

Post by Lianqing Zheng
Thanks a lot, Janne! I finally got it! Thanks for your inspiration! I
added 1-4 pairs in [pairs] and then got the right values. I thought
Gromacs would automatically find 1-4 pairs.

send the structure to pdb2gmx and the value of genpairs will be
determined from your ffXXX.itp file. As far as I know, this is only in
this way that the [pairs] section will be generated in the output.top
file. If your system is not a protein, then you need to add the paris
section (and in fact create the entire .top file) by hand.

By the way, why are you using the combination of genpairs=yes
fudgeLJ=0.5 and fudgeQQ=0.8333 ? This is not a valid combination for
any force field that comes in the standard gromacs distribution.

[Qiao Baofu]

Post by chris.neale
By the way, why are you using the combination of genpairs=yes
fudgeLJ=0.5 and fudgeQQ=0.8333 ? This is not a valid combination for
any force field that comes in the standard gromacs distribution.

Hi,

I am puzzled about this. In your opinion, when I use genpairs=no, fudgeLJ=
0.5 and fudgeQQ=0.8333, are these scaling factors, 0.5 and 0.8333, useful?
On the other hand, when using genpairs=yes, how to set the scaling factor of
fudgeLJ and fudgeQQ?$
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20061207/d8174d48/attachment.html>

[Lianqing Zheng]

Thanks, Chris!

I'm using AMBER FF, so those are the scaling factors in this force
field. Actually I never used standard Gromacs/Gromos ff and always make my
own top file from scratch.

Lianqing

Post by chris.neale

Post by Lianqing Zheng
Thanks a lot, Janne! I finally got it! Thanks for your inspiration! I
added 1-4 pairs in [pairs] and then got the right values. I thought
Gromacs would automatically find 1-4 pairs.

send the structure to pdb2gmx and the value of genpairs will be
determined from your ffXXX.itp file. As far as I know, this is only in
this way that the [pairs] section will be generated in the output.top
file. If your system is not a protein, then you need to add the paris
section (and in fact create the entire .top file) by hand.
By the way, why are you using the combination of genpairs=yes
fudgeLJ=0.5 and fudgeQQ=0.8333 ? This is not a valid combination for
any force field that comes in the standard gromacs distribution.
_______________________________________________
gmx-users mailing list gmx-users at gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-request at gromacs.org.
Can't post? Read http://www.gromacs.org/mailing_lists/users.php

[chris.neale]

Post by Qiao Baofu
I am puzzled about this. In your opinion, when I use genpairs=no, fudgeLJ=
0.5 and fudgeQQ=0.8333, are these scaling factors, 0.5 and 0.8333, useful?
On the other hand, when using genpairs=yes, how to set the scaling factor of
fudgeLJ and fudgeQQ?$

fudgeQQ is always used no matter what the value of genparis is.

(exerpt from
http://www.gromacs.org/pipermail/gmx-users/2006-September/023743.html)
Each force-field has its own rules (e.g. gen-pairs and FudgeLJ/QQ), but these
apply to the information outlined above. For example, gen-pairs does NOT mean
"generate a [ pairs ] section for the molecule." Instead, it means "If LJ-14
epsilon and sigma are not present in a [ pairs ] section entry, and
that type of
interaction is not explicitly formulated in [ pairtypes ], then it is
permissible to use the regular non-bonded parameters, and in that case scale
them by FudgeLJ."

Therefore your settings indicate that coulombic 1-4 interactions will
be scaled by 0.8333 and the pairs must be taken directly from the [
pairs ] or [ pairtypes ] section (and they will NOT be scaled by
fudgeLJ). In your case fudgeLJ does not matter (you could change the
value and it would not affect your simulation). However, I have always
hoped that it is set to 0.5 to indicate that this is what the
forcefield developers have done for you and included it in [ pairtypes
] so that it's kind of a reference value for your piece of mind.

I was assuming that the previous messages were from an Amber
forcefield. Since Amber uses (to my knowledge) [ pairtypes ] instead
of genpairs=yes, I was confused by the combination of genpairs=yes and
fudgeQQ=0.8333.