The following is the question of zhang_jaj:
Hello, I want to use towhee to do the nucleation process of argon gas on the surface under NVT, I hope you can give some guidance, first of all I want to choose potential as LJ, surface and particles are also LJ functions,
1) Question Does this surface need to be made with particles? If so, how should the particles that make it up be fixed
? 2) Or, do you directly use the LJ9-3 wall in towhee? I want to know how this wall works with particles, it Isn't it made of small particles?
3) After the simulation is completed, how to judge the structure of liquid and solid, is there any program code?
4) How to calculate the rate of this nucleation (it is the number of street-facing nuclei generated per unit time and unit volume), I want to know how to implement it in the program, I didn't find that parameter related, or do I need to add something myself? ?
5) How to implement this nucleation barrier?
6) Towhee cannot be parallelized, so in fact, a task is processed on one process; if my personal computer has a higher frequency than the server, is it faster to use a desktop computer than the server (because the desktop computer has 4 cores, Can it handle towhee's tasks by itself with 4 cores?)?
7) I hope you can give some experience about towhee code learning, which is for the direction I am going to do.
This is my answer: I
really can't do it, I'm still in the learning stage, because I haven't done a similar system, I can only have some suggestions for your problem.
1) The question is whether this surface needs to be made with particles, if so If so, how should the particles that make up it be fixed?
You can use the particle model to simulate, the fix is relatively simple, just don’t let these particles move in the move setting.
2) Or, can you directly use the LJ9-3 wall in the towhee? I want to know how this wall works with the particles. Isn't it made of small particles?
I haven't used this before, but there are people who do it in the literature. For the use of this potential energy, you can refer to the example Fris_Walls, and search the Internet. These posts and articles may be useful to you.
http://jcp.aip .org/resource/1/jcpsa6/v67/i5/p2384_s1
http://jcp.aip.org/resource/1/jcpsa6/v73/i8/p4050_s1
http://lammps.sandia.gov/threads/msg24356.html
http://www.sklogwiki.org/SklogWi ... ard-Jones_potential
This potential is derived from the transformation of particle potential energy, the fourth reference has a derivation process, the main advantage is that it can greatly improve the speed. The post of lammps follows Your situation is a bit similar. It is done using MD, which can be used for reference.
3) After the simulation is completed, how to judge the structure of liquid and solid? Is there any program code?
This is mainly to analyze RDF, the code is in utils, you can modify it according to your situation.
4) How to calculate the rate of this nucleation (it is the number of street-facing nuclei generated per unit time and unit volume), I want to know the program How to achieve it, I didn't find that there is any related parameter, or do I need to add something myself?
This is obtained by post-processing. Towhee itself does not calculate the nucleation process. It should be implemented by writing code. But how does MC calculate the time? If there is a problem with MC steps? Is it more appropriate to use MD?
5) How to implement this nucleation barrier ?
This should be done by simulating different temperature points, making a rate-temperature relationship diagram, and then getting it by fitting the thermodynamic formula.
6) Towhee cannot be parallelized, so in fact, one task is processed on one process; if my personal computer is more The main frequency of the server is high, so is it faster to use a desktop computer than the server (because the desktop computer has 4 cores, can it process the tasks of towhee by itself with 4 cores?)?
Towhee cannot use multiple cores to calculate a task in parallel, and in fact it does not need to. Generally speaking, simulating multiple tasks can be equivalent to simulating a large task. The nature of MC determines that it has such a nature, but MD cannot. So try to use the parallel mode provided by towhee. Pay attention to choosing different random seeds.
7) I hope you can give some experience about towhee code learning, which is for the direction I am going to do.
Regarding the code, I actually don't have much experience. I mainly post-process the results. towhee itself is a general Monte Carlo code, not specially designed for some systems, but it should be competent for your system. Mainly For the code of the system, the difference is mainly in the two processes of modeling and analysis results, and the core code is similar. I personally think that if time is limited, don't get stuck in the source code. Of course, you can learn more about the use of code for software It is also very beneficial.
The code of towhee is relatively clear because I need to read some of it. I can try to learn it in general.
For the post-processing code of the result, it is recommended to start by modifying the code provided by towhee. Take the one you are most familiar with, and if you can make do with it, fortran90 is recommended.
It is also very beneficial to learn more about linux bash scripts. I basically use it for simple data processing now.
Finally, I emphasize, don't get caught in the code, you can complete your own use, the simpler and more direct the better. Think more For questions related to your system, check the literature more.
/// I hope to learn more about towhee knowledge from you. The tutor doesn't care, and you can explore it by yourself. I'm worried about the top ten problems of graduation. Thank you.
Everyone is basically the same. You have firm confidence and you can do well. In fact, the direction of simulation calculation is more transparent than doing experiments. There is generally no problem in duplicating documents. Technical questions are generally answered on the forum. .I
am also in the process of learning, everyone discusses a lot and learns from each other.
————————————————————————————————————————————————————————————————————
Hello, I would like to ask you again. Regarding towhee parameter settings:
1) Are the values of .pm* arranged from small to large? One of the examples is
pmtraat 1.0
pmtracm 0.67
pmrotate 1.0
. When you want to judge which move to perform, should you only perform pmtraat?
2). When adding an LJ-9-3 wall to the system, is there a parameter in it that ljfrho (the number density of atoms in the integrated wall potential) is the point density on the wall that interacts with molecules in the system? Are ljfsig and ljfeps artificially specified sizes?
thank you very much!
————————————————————————————————
1. Yes, try with increasing probability. For example,
pm1 0.1
pm2 0.5
pm3 1.0
pm4 1.0
, then the probability of pm1 to 4 is 0.1, 0.4, 0.5, 0. As long as 1.0 appears, then the probability of the following pm will be 0.
2. No, the specific meaning and setting of the parameters depend on A couple of links I sent you.
————————————————————————————————————————
Hello, I also want to ask you a question about the hmatrix parameter. Mannul said that this coordinate matrix may not be orthogonal, but it must be a right hand rule. I'm thinking if other non-orthogonal coordinate matrices are used, will the program run much slower? ? ? (Because I personally think that it is much more convenient to calculate the distance between atoms in orthogonal coordinates (for example, I fix a single-layer graphene on one surface))
Another problem is that for the NVT system, if I first let N =1000, after T=50K, how to choose the size of the box (gas). At this time, different box sizes should correspond to different P. Is it using PV=NRT? Is the size of this box adjusted according to the formula I want P? , or an approximate box size in normal atmosphere?
thank you very much! ! !
————————————————————————————————————————————————
I haven't used a non-orthogonal box, but this should have little effect on speed. You can try.
For gas, you can use PV=NRT to estimate, pay attention to the unit. The pressure is specified according to your system requirements.
————————————————————————————————————————
There is also the ljfrho parameter in LJ-9-3wall. I don’t know if you are involved. I really can’t understand how this is calculated. Which atoms are used, divided by how much space (is it related to the scope of action)? Because English is very bad, I can't understand that sentence clearly, thank you very much!
————————————————————————————————————————
I looked at the parameter manual for the LJ93wall and lifrho refers to the number of molecules per cubic Angstrom. For example, if your wall is Si, the number of molecules in each unit cell is divided by the unit cell volume. If your wall is SiO2, it may be necessary to build two walls with overlapping positions, Si and O to build walls respectively. Such number densities are the number of Si in the SiO2 unit cell divided by the unit cell volume, and the number of O divided by the crystal. cell volume.
——————————————————————————————————————————————
Still don't understand this. According to mannul, the wall should be regarded as a general surface, and this ljfrho is the point distribution density on this surface. And the following sigma and epislon are the interaction parameters between the wall and the particles in the system. (How should this parameter be defined), and if the wall is regarded as Si and the like to calculate ljfrho, how should the arrangement of these Si atoms be determined (or is this a disordered surface)? I haven't been able to download the link to the article you gave before. I don't know if you have it. Can you give me a copy ( [email protected] )? Thanks a lot for the article about LJ9-3! http://jcp.aip.org/resource/1/jc..._s1?isAuthorized=no
————————————————————————————————————————————————
I started from the beginning.
We generally use all-particle potential energy such as UFF, amber, etc. in the simulation. Each atom of this potential energy acts as an independent particle and has its own force field parameters. In this way, the interaction energy of each pair of particles needs to be calculated during the simulation. If our system is the interaction of gas or liquid molecules and a surface, of course, we can also use the full particle potential parameter to simulate, but we can also integrate the force field parameters of all particles on the surface to make it a whole. 9-3 The wall potential energy is obtained from the spatial integration of the 12-6 particle potential energy. For the process of integral derivation, see: http://www.sklogwiki.org/SklogWi ... ard-Jones_potential , it is recommended that you can simply deduce it yourself. The advantage of this is that the number of particles is greatly reduced (the surface is treated as a particle with a specific shape), thereby reducing the amount of calculation. During the derivation process, the meanings of the original parameters sigma and epsilon in 12-6 have not changed, and they are inherited directly. In addition, a new parameter ljfrhho is added, which is related to the composition of the system, that is, the number of particles in the unit volume of the bulk phase.
I just use Si and SiO2 as an example, you have to get sigma epsilon and ljfrhho according to your system. What is the composition of the wall in your system?
It took so much effort to make it clear.
I don't have that document either, you can ask for help
——————————————————————————————————————————————————————
I feel that after you talked about it, you did have a lot of understanding of this situation (I was very confused before). I want to try to simulate the nucleation process of Ar using single layer graphene and double layer graphene. The a of this 2D, such as a single layer, is 2.46 (Angstrom), so when I calculate it, should it be 2/(2.46*[sqrt(3)/2]*2.46)=0.38....
Another example is an fcc (111), adding his a=2, should the calculated value be 4/(2^3)=0.5? In this case, it has nothing to do with which aspect of fcc is selected (or I still understand it wrong 
)?
Also, I thought that adding a LJ9-3wall (such as graphene) was a virtual wall; but if I add another one of the same composition, such as graphene-wall, in other locations, and give the coordinates of the atoms in this graphene, and This wall is not the potential of LJ9-3. In other words, as long as a surface for adsorption is added in towhee, should it be treated as LJ9-3wall? In this way, his atomic arrangement is still uncertain (for example, when simulating epitaxial growth, different planes must be selected)?
So adding a self-introduced surface should be completely different from adding an LJ9-3, right?
Also, based on your experience, you think graphene has nearly 1000 carbon atoms, which should be used as an independent macromolecule to create the initial file. Which way do you think should be used? The manual parameter input of 1000 functions of vibration is too scary.
I've been really unsure about these issues recently, so I can only think of bothering you.
——————————————————————————————————————
Another problem is that I used ms to build the graphene model and export the pdb file format, but it is in the ATOM bit, and the pdb2towhee conversion seems to only be HETATM. This format conversion is a problem. Should I write a script for the conversion myself?
There is also the problem of nucleation of Ar gas on graphene. I put Ar in towhee_coords, and then put graphene in towhee_tempalte. This graphene uses the force field of trappe-ua. The problem is that each carbon atom has To enter a parameter of vibration into a key, I feel that the workload is particularly heavy. I don't know if graphene has a better input method. Thanks!
——————————————————————————————————————————
The 9-3 Wall potential simulation may not be suitable for single-layer and double-layer graphene because the surface thickness is assumed to be infinite during the integral derivation. 1000 C atoms using all-atom potential is also acceptable.
Your understanding of the density calculation is still not quite right. 9-3 The thickness of the surface is considered to be infinite when integrating the wall, then its density is equal to the density of the bulk phase, which has nothing to do with the cut surface.
I don't understand what your second paragraph means. Can you tell me more briefly?
The modeling of macromolecules in towhee is really a headache. However, if your graphene is stationary during the simulation, it's easy. You treat 1000 atoms of graphene as 1000 individual molecules, so you don't need to enter the vibration parameter.
————————————————————————————————————————————
If there is a problem with pdb conversion, pdb2towhee.f90 can be modified, or it can be implemented by scripting
——————————————————————————————————
Thank you very much, I went to treat each atom of this macromolecule as a separate molecule this afternoon, and the program ran successfully. I thought about it, in fact, each c atom is treated as an independent body in the force field, so it is also feasible to treat it as a lot of song molecules when it is fixed (give me fun in the afternoon, hehe).
In fact, I found this LJ-9-3 to be a diffuse surface. What I said in that paragraph was mainly to determine the atomic arrangement on the surface. Now I figured it out, it should not have this atomic arrangement. concept.
I hope you can give me an example. For example, I use the iron of bcc and fcc to make this LJ-9-3 wall, and their lattice constants are all 2 (Angtrosm), so how do I get this ljfrho respectively? , how much is it worth?
As for the values of ljgsig and ljfeps below ljfrho, is it necessary to rebuild a model to perform a cycle with the potential of Fe and the potential of the system, and then see the sigma and epsilon values of Fe and the atoms in the system from the towhee-out?
There is also the introduction of lj-9-3 in towhee mannul. The sigma and epsilon in the following formula should be a series of ljfsig and ljfeps values under ljfntypes. In this case, if the system has n kinds of atoms that interact with the surface, then in the integral In the process, there should be the integral formula of lj-9-3 in n (the parameters are different)
, I also got the pdb2towhee, use ":1, $s/ATOM /HETATM /g", and then you can.
————————————————————————————————————————————
Progress is worth celebrating.
For example, for iron in bcc, the volume of each unit cell is 2^3=8, and it contains 2 Fe atoms, then its ljrho should be 2/8=0.25.
ljgsig and ljfeps you are right.
The most important thing for learning software is to learn from manuals. Towhee's manuals are quite detailed. I also read them one by one at that time, and then combined with examples to learn.
——————————————————————————————————————————————————
Regarding modeling involving graphene, since graphene is fixed, I modeled each atom as a separate molecule in the way you said. There are problems in this:
1. Since the number of molecules suddenly increases, the three values of inix, iniy, and iniz need to be increased, but I don't know what these three values mean. Does it make sense to divide the box (because atoms The coordinates are all read from towhee_coords)?
2. When I test with c60, this method works fine, but when I use this method with graphene involving more than 1000 carbon atoms, everything works fine, but there is a problem in the output file: the information is only The output about the interatomic potential is gone, that is, the LJ formula is displayed, and then the sigma epsilon column, there is nothing below. . . I don't know what's going on, but the program is still running, and the box_*.pdb files are constantly being generated. I'm so puzzled, I
hope you can help me troubleshoot, thank you
————————————————————————————————————————
I tried it in the afternoon, and it seems that because it is not finished, it will not be displayed there. . .
——————————————————————————————————————————————
Hello, landlord, I have just come into contact with the towhee software. I am doing the gas-liquid balance of the ternary extraction system. Now I need to simulate the gas-liquid balance of the pure component thiophene. Before the simulation, I looked at the In the force field file, I found that some of the bond length and bond angle files I need are missing. Does the force field file not display all the bond length and bond angle data that can be processed? (Because I read a document before saying that the Trappe-EH force field of the towhee clock can handle the problem of N atoms, but after I entered the force field file of the towhee 7.0.2 version, I found that there was no data for N atoms at all. So There is such doubt ##).
If there is really no data I should add the force field myself? Is it necessary to simply find the N atom data from another joint atomic force field and paste it into the one I am using now?
Younger brother, I'm sorry to bother you!
——————————————————————————————————————————————————
I am not very familiar with this force field. The best way is to find the parameters of Trappe in the literature, and then add them to the force field file according to the format of towhee. If the Trappe force field does not have the parameters you need, you can get the parameters from other force fields, and make sure that the two force fields have the same form (for example, both LJ 12-6).
————————————————————————————————————————————————————
Hello, landlord, I have a few more questions to ask you.
1. The mixing rules of the two force fields are different. Is there any way to introduce a mixing rule for one of them? What if it could be done?
2. When manually entering the charge, if there is a charge value under the Coulombic interactions item in the description file of the force field, should I enter the charge value? If there is no value in the Coulombic interactions, should I not enter the charge value? , which is 0.0d0? Into the Dubb2004 potential, when used to build SiO2 molecules?
3. When creating a molecule, if there are multiple atoms of units, then in the towhee_coords file, should one type of atom be input according to the unit value under input_style and then another?
4. Do you know how to judge whether the material is fcc or bcc in the result file pdb or movie file of this program, or do you know which software has this function?
5. (Supplementary) For different force field files, for example, the epsilon value of argon is different. Does it mean that argon has different lattice constants for different force fields when it is solid?
6. For those involving isotopes, let’s take H and D as an example. If I have H’s force field file, can I modify H’s mass and use it as D’s position file? (for LJ potential)
Thank you so much!
————————————————————————————————————————————————————————
1. It is generally not recommended to use in combination with different mixing rules. You can test it to see if the impact is obvious.
2. The charge can be written as 0 without considering the Coulomb interaction, and the Coulomb interaction can be turned off if all are not considered
3. Don't understand? ? for example?
4. This needs to be judged by yourself, right? I have not found such a function.
5. It should be like this, pay attention to also compare the form
of the force field 6. Yes.
————————————————————————————————————
Hello:
1. I don't know if different mixing rules can be used, so it is useless. I can try it in the afternoon
. 2. That is to say, as long as my system does not want to test the Coulomb effect, I can write no charge. It's worth it
3. The problem is solved, because si and o in the pdb file of sio2 that I exported were written at intervals. For towhee, I need to write the coordinates of si and then write the coordinates of o.
4. Because I am doing a low-temperature crystallization process of gas, I need to judge what kind of structure is frozen. This is very depressing
. 6. I know that this can be modified for vasp, but for the force field file of LJ potential, Because I don't know whether the parameters are related to the mass of the atom, I don't know whether the isotope force field obtained by this modification can be used.
Question: I also want to ask, do you know the nucleation process? I don't know how to calculate the nucleation barrier of the system during the simulation process. If the total energy of the system during nucleation is subtracted from the initial total energy, does this potential barrier depend on the atomic arrangement of the initial model (because I read the files generated by packmol in your tutorial, so I don't know if this generated arrangement is general when calculating the potential barrier.)
Are there two kinds of sampling, umbrella sampling and flux forwar sampling, in towhee? If you don't want to add it, should it be added to the compiled program, or should it be recompiled after adding it?
thank you very much! ! ! !
——————————————————————————————————————————————————
6. For isotopes: Quantum mechanics can all use the same parameters, and molecular mechanics can of course also do
n't know much about nucleation. The nucleation barrier can be reduced by multiple simulations to reduce the influence of the initial structure.
Umbrella sampling is available in towhee, the latter one is not very clear, you can look at the manual.
If you want to modify the program, of course you need to recompile!
——————————————————————————————————————
If the force field file can be directly modified as an isotope, then if I need to involve the mixing of H2 and D2 gases, do I need to modify the force field that copies a H2 and then rename it or put it in another directory What about calling?
————————————————————————————————————————————————————
You can do this, or you can add the D atom type to the original force field
——————————————————————————————————————————————————————
Hello, I have some doubts about the isotope force field file, if you change the value of mass, it can be used. If it says about H and D, the mass is different, and then everything else is the same (force field), then the two simulations going in separately should behave the same? Because mass seems to have no effect on the potential energy in LJ. In this case, the isotopes are actually inseparable, right?
————————————————————————————————————————————————————————
LJ potential energy really has nothing to do with mass. But the whole process of the simulation is like this:
initial velocity and coordinates ---> find the force according to the potential energy function --> find the acceleration according to Newton's second law (!!!, related to mass) --> find the speed from the acceleration ---> Find position from velocity --> ...... loop
————————————————————————————————————————————————————————————
This is molecular dynamics, right? If it is Monka, then there is no way to distinguish H2 from D2, right?
————————————————————————————————————————————————————
Sorry for the confusion.
I think there should be no way to distinguish
——————————————————————————————————————————————
That's stupid, is it only possible to change lammps? 
I didn't see D2 and T2 in the force field in towhee. .
——————————————————————————————————————————————————————
I don't know whether you are studying the kinetic effect or thermodynamic effect of isotopes. If it is a kinetic effect, MC itself cannot do it. If it's a thermodynamic effect, to be honest, I don't yet understand the difference between them
——————————————————————————————————————————————————————————————————————————
I want to do a simulation of the crystallization process of H2 and D2 at low temperature, see their nucleation process, etc., but as far as I know, their triple points are significantly different, so the temperature of gas liquefaction is also different. But if you use the same force field, it will feel the same. . . Can't show any difference?
——————————————————————————————————————————————————————————————————————————
Hello, it seems that I can only trouble you again. You told me before that there is an algorithm of umbrella sampling in towhee, but I will not use it to determine the potential barrier of my nucleation process. I don't know how to do this. Could you give me a general process? Thank you very much, and I don't have to worry about that potential function. For now, I won't let you do that, only argon.
——————————————————————————————————————————————————————————————————————————
Hello, I have to trouble you again!
I discussed building a graphene model with you before. At that time, all C atoms were treated as independent molecular, that is, the potential function only considered non-bonded parameters and did not consider bond parameters. The premise of this is that graphene does not participate in the move (in this case, it is better to make the CC effect 0, and I still don’t know how to do it. Is it necessary to obtain the C-Ar parameters from the geometric method alone to make a force field?).
But recently a senior brother said that my model is too ideal, and the C atom should also be moved. If this is the case, the non-bonding effect in graphene needs to be considered. If this is the case, I don't know how to build this graphene model?
If you count all the keys, you have to lose too many lines below towhee_input. . .
On this issue, senior, what do you think?
———————————————————————————————————————————————————————————————————————————————