Topic Options
#15336 - 08/26/07 12:34 PM Basic questions about IMAGe and CRYStal
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
Dear all,

Though I read the documents of images.doc and crystl.doc and some testcase scripts, I still have some confusions about the two modules:
1>. how to set up CUTOFF(in CRYStal build CUTOFF) and CUTIm(in image updating)? What's the relationship between the two parameters?

2>. How to know the number of Transformations by defining: CRYStal BUILd cutoff ?
For example, if a system spans a range: Xmin=-5, Xmax=5; Ymin=-4,Ymax=4; Zmin=-3, Zmax=3
and followed a command: CRYStal BUILd CUTOFF 10

3>. What are the meanings in output file(in red)?
i>: Range of Grid Search for Transformation 1
Lattice Vector A -3 TO 3
ii>:The number of transformations generated = 208

Number Symop A B C Distance
1 1 -1 -2 -2 21.4577
iii>:Transformation Atoms Groups Residues Upper-Bound
9 C009 has 0 0 0 24.36
4>. If we have a system spans a range: Xmin=-10, Xmax=10; Ymin=-10,Ymax=10; Zmin=-10, Zmax=10 and a command: CRYStal DEFIne ORTH 7 7 7 90 90 90, what will be the results?

Are the parts of the system beyond the ORTH range neglected during the following calculations?


These questions may seems stupid, but it will give me a profound understand if I can know them clearly. Any response is greatly appreciated.

Thanks.

Top
#15337 - 08/26/07 01:33 PM Re: Basic questions about IMAGe and CRYStal [Re: jeffrey]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8373
Loc: 39 03 48 N, 77 06 54 W
The condition CUTOFF > CUTIM must be the case, perhaps CUTOFF >> CUTIM; it depends on how well the volume of space matching the unit cell is filled with atoms. CRYSTAL BUILD uses existing atoms and CUTOFF to determine which image transformations to define; if the volume is only partly filled and CUTOFF is too small, some needed transformations might be omitted. A reasonably safe value for CUTOFF is about half the longest unit cell dimension (unless that is smaller than CUTIM).

The number of transformations defined by CRYSTAL BUILD is printed in the output; the number required depends on the lattice type. For six-sided parallelepipeds (CUBIC, TETR, ORTH, MONO, TRIC) a minimum of 26 transformations are needed. The minimum number of transformations needed for other lattice types (HEXA, OCTA, RHDO) is shape dependent.

The lattice vector reported is the number of unit cells in each direction evaluated for possible image transformations within CUTOFF of any existing atom.

The numbers are distances from existing atoms to image transformations of existing atoms.

For a 10 A cube filled with matter, imposing a 7 A CUBIC unit cell would probably create a lot of badly overlapped atoms. You should compute the density, based on the mass of the system and unit cell volume. The maximum extents (XMIN, XMAX, etc.) do not indicate the distribution of mass, and have a very approximate relationship to the unit cell size.

Long range effects are usually included via the use of P-M Ewald summation for electrostatics; there are several approaches to long range VDW forces, depending on the ensemble and CHARMM release.

Top
#15338 - 08/27/07 04:33 AM Re: Basic questions about IMAGe and CRYStal [Re: rmv]
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
Hi Rick,
Thanks very much for your response. It indeed helps me much to understand the two modules. But I still have something unclear about that.

1.What's the purpose to invoke the CRYStal function in a md run? To set up the periodic boundary condition? If so, the IMAGe can also do it, doesn't it?
2. What is the results after the commonds CRYStal DEFIne and CRYStal BUILD? Is it a crystal, consisting of the primary atoms and their images, which has the type and parameters as defined in CRYStal DEFIne?
3.For example, a peptide is solvated by a water box having dimesions 50 x 60 x 70A, what is the best choice of the crystal? is it defined by CRYStal DEFIne ORTH 50 60 70 90 90 90?
4.Does the origin of the coordinate locate at the center of the crystal defined?
5. I found that the number of transformations in testcase c25test/allxtl.inp is 16 for CRYSTAL CUBIC, not the least number 26. Which reason may result in such a difference?
---------
Crystal Parameters : Crystal Type = CUBI
A = 20.00000 B = 20.00000 C = 20.00000
Alpha = 90.00000 Beta = 90.00000 Gamma = 90.00000

CHARMM> crystal build cutoff 16.0
XBUILD> Building all transformations with a minimum atom-atom
contact distance of less than 16.00 Angstroms.

Range of Grid Search for Transformation 1 :
Lattice Vector A -2 TO 2
Lattice Vector B -2 TO 2
Lattice Vector C -2 TO 2


The number of transformations generated = 16


Number Symop A B C Distance

1 1 -1 0 -1 15.7942
2 1 0 0 -1 8.3550
3 1 0 1 -1 15.2904
4 1 -1 -1 0 15.8631
...
------------
Maybe, for some questions, I have a wrong origin of my deduction, causing so many confusions. So any suggestion is appreciated.

Thanks in advance.

Top
#15339 - 08/27/07 02:08 PM Re: Basic questions about IMAGe and CRYStal [Re: jeffrey]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8373
Loc: 39 03 48 N, 77 06 54 W
CRYSTAL uses the underlying IMAGE structure, but automates the generation of transformations, which can be quite complicated for some of the lattice types. In addition, some newer features in CHARMM such as Ewald summation and constant pressure dynamics require the use of CRYSTAL (as opposed to IMAGE alone).

CRYSTAL BUILD creates the image transformations.

For solvated molecules, CUBIC, OCTA or RHDO are recommended, rather than ORTH; two principal reasons are: [1] molecules can rotate during MD, and [2] for constant P, an ORTH box can change shape.

The default origin is 0,0,0 and this is typical, but there are some exceptions.

There may a bug of sorts in the testcase; it looks like only a single water molecule is used, so "CUTOFF 16" may miss some transformations for a 20 A cube. Testcases are not always good models for user input.

Top
#15340 - 08/28/07 02:09 AM Re: Basic questions about IMAGe and CRYStal [Re: rmv]
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
Thank you very much,Rick.
I will reread the documents and testcases relating


Edited by jeffrey (08/28/07 02:10 AM)

Top
#15341 - 08/28/07 11:43 AM Re: Basic questions about IMAGe and CRYStal [Re: jeffrey]
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
Dear all,

I did a testing system consisting just two TIP3 H2O to demonstrate the operation in the process of CRYStal DEFIne & CRYStal BUILD. Enclosed are the output file and primary+image coordinate file.(In order to get complete image of the primary atoms, I set CUTIM larger than CUTOF) But some confusions still exsit.
1. I still don't know how to calculate the first three "Distance"s appearing in the following output sentence. It seems that the 4th distance 10.0000 is the distance of transformation of the primary atoms along C-vector.
---
Range of Grid Search for Transformation 1 :
Lattice Vector A -2 TO 2
Lattice Vector B -3 TO 3
Lattice Vector C -3 TO 3

The number of transformations generated = 44
Number Symop A B C Distance
1 1 0 0 -2 13.9188
2 1 -1 0 -1 13.1154
3 1 0 -1 -1 13.7549
4 1 0 0 -1 10.000
...
-----------

2. For cubic crystal, is the origin of the coordinate at the center of the cube?
(hmm, the way to ask this question may be wrong... Or this question may be nonsense,because we just transform the system as a certain type of crystal and the cutof value, having nothing to do with the origin of the coordinate system)

3. Should the model(for example, solvent+peptide) I build for MD simulation have the same density as the value used in experiment? If so, how to calculate it with certain type of crystal? Should image atoms and volumes be contained in the calculation?

4. For a properly built model, the primary atoms and the nearest images should be adjacent, without any vacancy between them. Is it right?

Any suggestion is greatly appreciated.

Thanks.


Attachments
15360-output.txt (283 downloads)


Top
#15342 - 08/28/07 11:44 AM Re: Basic questions about IMAGe and CRYStal [Re: jeffrey]
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
coordinate file for primary and image atoms


Attachments
15361-coorpdb.txt (300 downloads)


Top
#15343 - 08/28/07 01:21 PM Re: Basic questions about IMAGe and CRYStal [Re: jeffrey]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8373
Loc: 39 03 48 N, 77 06 54 W
I've never bothered to check the exact details of that distance calc myself; you may want to go source code diving if you're curious.

As for the 10.0 for the 0 0 -1 lattice vector, one of the water molecules has 0.00 for all Z coords, the other has positive Z coords, so in the -Z direction 10.0 would appear to be the distance to the closest image atom in that direction.

For most MD applications using simple P1 symmetry, the lattice is centered at the Cartesian origin (0,0,0).

You should definitely attempt to match expt density, using the volume of the unit cell; the volume formulas for each lattice type are given in crystl.doc. Only primary atoms should be included. For a simple cube

set a 10.
calc v = @A * @A *@A
calc rho = 1.673 * ( ?MASS / @V )

where 1.673 is a conversion factor from amu/A^3 to g/cc and is based on the proton rest mass.

A well built model has no overlapping atoms, nor any void volumes.

Top
#15344 - 08/29/07 01:37 AM Re: Basic questions about IMAGe and CRYStal [Re: rmv]
jeffrey Offline
Forum Member

Registered: 03/18/07
Posts: 148
hmmm,I have a better understanding now.
Thanks so mcuh, Rick.

Top

Moderator:  John Legato, lennart