chain.xyz file attached below. Although this file is extremely simple, it has two problems when loading it into VMD. Its atom type "1" will be interpreted by VMD as hydrogen and thus the automatic bond search will not work as its heuristics do not allow for bonds between hydrogen atoms. Furthermore the atom radius associated with hydrogen atoms (or other "normal" atoms) is to too large for the bond search to work correctly. We'll have to address these issues before telling VMD to guess the bonds and then generate additional topology information based on those guessed bonds.
This paragraph discusses the simplexyztodata.tcl script. The first block of script code loads the TopoTools and pbctools plugins and makes sure that they are of a sufficiently high version. This test makes it look a bit more complicated, but effectively this is the same as these two lines:
The next step is to load the molecular coordinate data. We use the
Now we need to assign data to the atoms that is required for writing a proper LAMMPS data file, but not available in the .xyz file. This is done by creating an atom selection and then assigning properties to all atoms in that selection:
This specific system is to be used with reduced units (units lj in LAMMPS input) and thus we need to set a reasonable radius and mass value. This now is sufficient to compute the bonds from atom distances. The algorithm implemented by VMD is simple: if the distance between two atoms is shorter than 0.6 times the sum of their radii, those atoms are considered to have a bond. In the example coordinate file, that atoms are 1.0 units apart and we have set the radius to 0.85. Thus 0.6 * (0.85+0.85) = 1.02 which is larger than 1.0, so there will be a bond between each atom and its direct neighbor(s):
This step does not assign any bond type (or bond order) information, but rather define it as "unknown". In many classical MD packages, the bond type is derived from the atom types that form the bond. TopoTools contains a script that helps with assigning bond types according to this convention. In this specific system, all atom types are the same, so there should be only one bond type:
Now all bonds have been given the type "A-A" based on being between bonds of two atoms with type "A". The force field used for this example also requires the definition of an angle potential between neighboring bonds. This is done via:
This step defines an angle between each triple of atoms that consists of two bonds sharing one atom. The angle typing is done in automatically in this step and in a similar fashion as in the bond typing step, i.e. all angles should have an "A-A-A" symbolic type.
As a final step we ask VMD to re-analyze the molecular data since we have changed bonds and angles and we need to update any data that VMD derives from that information, e.g. which atoms belong to a single molecule.
Finally we need to set the simulation cell dimensions and write out the final state as a data file for the desired "angle" atom style:
The last script command
The attached script file in.step1a is an example for how to use the resulting data file in a LAMMPS calculation. This input will do a very short MD run with random velocities to randomized the coordinates and then request a minimization. The restart written at the end of of the calculation can be converted with the restart2data tool to a data file, data.restart-1a, and this can be read into VMD with:
Now we want to look at a model polymer that has different atom types at the beginning and end of the chains and we discuss the twotypesxyztodata.tcl script. This is similar to the previous one, so only the differences are explained. We use the same coordinate file and most of the initial steps are the same, but we need to define a second selection to change the properties of the two atoms at the beginning and end of the 11-atom chain (which are listed in VMD as index 0 and 10, respectively):
The bond detection also works the same, however, after retyping we should get two bond types, "A-A" and "A-B" instead of one. A list of bond types can be retrieved through:
This will also indicate the order in which those symbolic bond type names will be mapped to numbers starting from 1 in the data file. This output of the script indicates that bond type 1 will be set for bonds between one atom of type A and one atom of type B, which bond type 2 will be assigned for bonds between two atoms of type A. Similarly, we'll get two angletypes, "A-A-A" and "A-A-B", respectively.
The remainder of the script is equivalent to the previous one, however, to run the calculation in LAMMPS some more parameters will need to be defined, so a new input script, in.step1b, is required.
In this example we will read in the same data file as before, but want to create a topology that has two molecules instead of one, with one of them being translated relative to the other. The corresponding VMD Tcl script, twochainsfromone.tcl, starts like the script in step 1a, but after reading in the coordinate data and assigning its properties, we need to define a new "molecule" in VMD that is able to hold all coordinate information, since VMD does not allow to change the number of atoms in a data set after it has been created. Please note that in "VMD-speak" a "molecule" refers to the whole data set, regardless of how many actual molecules in contains. The animate command reserves storage for coordinate data, which is not automatically allocated unless coordinate data is read from a file.
Now we collect all information from the old molecule that we want to transfer to the new molecule in a list.
Then we change the original molecule's coordinates by translating it 4 distance units in x-direction:
Finally we create a new selection for all new atoms, join the two lists of atom properties and assign them to the atoms in the new molecule:
From here on the script is equivalent to the previous ones and the MD input in.step1c is essentiall the same as in.step1a, except for the changes in file names, e.g. for the
In some cases the distance based automatic bond detection may fail like shown in the top picture on the right which was generated by using the script from step 1a on the spiral.xyz file. The script spiralchaintodata.tcl shows an example for how to overcome this kind of problem. This is a system where the particles are arranged in the shape of a spiral (cf. second image on the right), but towards the center the atoms from different turns are closer than the atoms that are bonded together. In this case we have to search for and alternate way to construct the list of bonds. For the system at hand we know that all atoms for a chain according to their order in the coordinate file. With that additional information we can assemble the list of bonds manually with the following VMD Tcl code:
Since there is only one type of atom, we also set the bond type to A-A directly. If some atom types would be changed, the bond types can always be updated with topo retypebonds. The rest of the script code is like before with one addition. In order to easy visualization we write out the .psf file, that can be read in at a later point to reset the bonding information to what we have computed with out script:
The initial coordinates have some pretty high potential energy, so the MD input in.step1d has to be a bit more complicated to get via a few quenches (running some MD followed by minimization) to a somewhat relaxed structure.