Activity for Marcus Martin
-
Marcus Martin posted a comment on discussion Help
Do you have a copy of that paper or at least a reference for me to look at? I am not familiar with those terms, but I am a bit out of the loop on the literature these days. Marcus
-
Looking at the pdb files can be deceiving as you have some molecules wrapping periodically through the box so that can make it look like the molecules are outside of the box, when really they are wrapping around to the other side of the box (each molecule has its center of mass in the box, but the atoms may stick out). That picture of the system after 20000 moves does look strange. If you include all the input and output files from the run I have a much better chance of figuring out what is going...
-
Looking at the pdb files can be deceiving as you have some molecules wrapping periodically through the box so that can make it look like the molecules are outside of the box, when really they are wrapping around to the other side of the box (each molecule has its center of mass in the box, but the atoms may stick out). That picture of the system after 20000 moves does look strange. If you include all the input and output files from the run I have a much better chance of figuring out what is going...
-
There is not currently an option for cylindrical unit cells. That isn't a shape that can be periodically repeated to fill space so I am not sure how that would work. If you have a reference to something that describes how that unit cell would function I can take a look at it and estimate how much work it would be to implement.
-
Howdy Jessie, Looks like a constant volume cube of 63 by 63 by 63. What evidence do you have of the x, y, and z values fluctuating? Marcus Marcus G. Martin 88 Martinez Road Edgewood NM 87015-8222 (505) 363 3179 www.photobirder.com towhee.sourceforge.net
-
Howdy Jessie, I do not see anything that would be changing the box size, especially given this is a grand canonical simulation. Can you include the output from that simulation so I can check to see that the box values are really changing and not just something else related to x,y,z (like maximum translation values)? Marcus On Wednesday, March 22, 2023, Jessie jessie20230322@users.sourceforge.net wrote: Hello everyone, I'm trying to use GCMC to simulate the nitrogen behavior in the bulk phase at 77.4k....
-
Howdy, They are processed at the same time. If you run a list of 4 jobs on 4 nodes then it is 1 job per node until they are all done. Marcus On Tue, Aug 23, 2022 at 3:21 AM wangqianqian wqq@users.sourceforge.net wrote: Thanks very much! 1. So, I understand that the MPI version of Towhee which reads from an additional file (towhee_parallel) is not able to be processed by multiple processors, right? 2. In fact, the file(towhee_parallel ) is a batch file, when I use the option towhee -p jobfarm to process...
-
Pretty much the same, maybe a tiny amount more overhead. Using 4 processors to compute a single task is just 3 processors waiting around while the other one runs the simulation. Perhaps slightly slower due to how your machine communicates. Really want to have more (or equal) jobs than processors for jobfarming.
-
version 8.2.3
-
version 8.2.2
-
I could not reproduce this error. That command worked for me on my development machine. Were you using the mpi compiled version of the code (enabled with the -enable-mpi configuration flag prior to compiling) when running this command? Strange that there is screen output in this portion as when using the MPI compiled version of Towhee that gets sent to a different file.
-
Bug addressed in version 8.2.2.
-
Bug fix for this issue addressed in version 8.2.2.
-
Howdy Even, I put very little effort into constructing a starting configuration for my simulations since it should not matter much once the system starts running and sorts everything out. Unless you are setting up a molecule that will not move (like a porous solid) you can in theory start from anything and get to a good configuration by just doing some of the simple translation and rotation moves. You have encountered the one time when that does not work, namely when there is a hard overlap. While...
-
Thanks for that bug report. Other people had similar issues lately so I had already been working on a fix. The configure file was modified to work better with Intel compilers and I also fixed that mistake where I went over the line limit with the preprocessor options. Version 8.2.1 is now available and should resolve your problems. If you see anything else please let me know. I do not have a machine setup with the combination of operating system and compiler that is giving folks trouble (seems like...
-
version 8.2.1
-
Howdy Rafie, Did some debugging now that we have the files sorted out. Here are the angles computed for the 4 water molecules in your system (in degrees) imolty,units,angle,: 1 3 1 2 104.39856834930609 imolty,units,angle,: 1 3 1 2 104.46425669231122 imolty,units,angle,: 1 3 1 2 104.53565304465687 imolty,units,angle,: 1 3 1 2 104.29203343601303 That last one is outside of the tolerance allowed for the rigid angle by the default TIP3P-EW settings. To be honest, I was surprised that the other 3 were...
-
The towhe_ff_GO file you provided appears to be a copy of the towhee_ff_TIP3P-EW file and so I get error messages related to not finding the parameters for the atomtype 'CA'. If you can try to provde the correct version of the towhee_ff_GO file I can take a look at the problem in more detail. From looking at your towhee_output file the error is happening for the first water molecule in your simulation. That is using a rigid angle so I assume the angle computed from the first 3 atoms must not be quite...
-
Looks like you were also using your own forcefield files so I cannot run this without them. I did take a look at the towhee_output and have some comments that might help. The error means the first chain in the system, which is molecule type 1, has a problem for the bond between atom 2 and atom 1. That means the first water molecule in your simluation, and the bond between the oxygen and hydrogen. The bond length is computed as 1.35974262270475, and it is bond type 1 in the simulation. If I go back...
-
What I meant to convey was there is no theoretical maximum allowed value of NUMAX, just however big you need. Clearly that is confusing so I will think about how to reword some of that section. You can find preproc.h in the Source directory. Just search for the NUMAX in there and set the value that follows it to something slightly larger than your bigest molecule. Marcus
-
Yes, you need to make sure NUMAX in preproc.h is larger than the biggest value of nunit in towhee_input. You will then need to recompile the code with that new value. The only comment I could find about NUMAX in the basic connectivity map manual at http://towhee.sourceforge.net/inpstyle/inpstyle_2.html was nunit (integer) The number of atoms (or united-atoms) in this molecule. Must be less than or equal to NUMAX. Where did you see the comment about no maximum number on NUMAX? thanks, Marcus
-
Trying to narrow down why you are seeing different results. Here are some things I noticed about your Towhee simulations 1) You are spending 25% of the moves on performing rotations. Those moves do not do anything for single-atom molecules like Ni and H. That is why there is 100% acceptance rate as rotating an EAM atom does nothing to the simulation. You would be better served spending that time on Translations. 2) The simulation has not finished equilibrating during this first run. You can tell...
-
No. From a theoretical point of view I believe a system is in its lowest energy state at 0 K so not completely sure an ensemble of states has any useful meaning. From a practical point of view, all of the enegy states are scaled by 1/kT so zero temperature results in dividing by zero. If you are just trying to do a minimization I suppose you could set a very low temperature, but Towhee is not well suited for minimizations so the results are almost certain to hit a local minimum instead of a global...
-
Intake of Hydrogen increasing beyond your reservior is more a problem with the simulation having more absorption that you expected as opposed to a problem with Towhee. Clearly you need to set the value higher until there is some left over that is not absorbed. For the chemical potential conversion there is an EV to Kelvin conversion factor in Towhee of CON_EVTOK = 11604.449d0 that might help you convert your units. Marcus
-
Yes, the order of the atoms in your towhee_coords file should match the order of the atoms in towhee_input. The towhee_initial file is only used when linit is set to false. If you are reloading from a towhee_coords file after running in MD it is mostly the same as starting a new simulation (linit set to true) but you are correct that you will want to make sure the box sizes match up properly as that is not read from towhee_coords. Marcus
-
The Lennard-Jones potential, like the majority of van der Waals type potentials, has an attractive terms that scales as 1/r^6 and that continues to create some attractive forces/energy all the way out to infinity. Correspondingly, the repulsive 1/r^12 term also has a small contribution out to infinity, but it scales down substantially faster than the attractive part. The two way of handling long range interactions are 1) Just scale the potential to be zero at the cutoff and define it to be zero past...
-
preliminary 8.2.1
-
Very likely you already found the problem and that is doing insertions of large chain molecules into a fluid is difficult. The new chemical potential distribution tools added in version 8.2.0 might help break down the issue. For best sampling, we would be getting a nice bell curve of chemical potentials centered around the reported mean. More likely, the sampling is poor and we have a lot of zero results from failed attempts, and a handful of nonzeros. In that case we really need to improve the sampling...
-
You are running using an older towhee_input file and it is getting a bit confused and skipping past all of the move information. It then complains because without any move information, it doesn't see you having a complete set of probabilities for performing Monte Carlo moves. If you are still having problems like this let me know and I can help you get your files updated for the current version. Marcus
-
Since the box has periodic boundaries the effect will still be in the box, but I understand how it is more satisfying to have all of the coordinates in their proper location. It seems like what you are doing should work, so in order to figure out what is going wrong I would need access to the towhee_input file and any other files you are using to set up the simulation. If you are still having this problem and want to send me those files I can take a look at it. Marcus
-
schedule.F is the routine that sets up all of the logic for CBMC growth patterns. rosenbluth.F is the main routine for running a CBMC growth. The details of how to generate trials during that procedure depend on the many customizable options for the CBMC moves, most of which are hidden behind some defaults as the majority of the time people do not want to know the details and just want something that works. A combination of the Towhee manual, and some of the papers I have written about the details...
-
I am a biased source to ask about chemical potential because I have found it not to be trustworthy in many situations over the years so I try very hard to use ensembles where I do not need the value of the chemical potential and just compute it as an additional check on whether things look equilibrated. For any of the larger molecules you will hit a point where the density is sufficiently high that the chemical potential becomes very noisy. Even when using the Gibbs ensemble you have a related issue...
-
I think you have an extra space in the Atom Names in the bonded section of that file. Should be a10,1x,a10 (so 10 characters for the name, one space, and 10 characters for the next name). Looks like yours has an extra space in between which probably means the code is reading it in as ' HW' instead of as 'HW'. Would need the towhee_input file to debug it for sure. Marcus
-
Version 8.2.0 now has optional parameter loutchempotdist (default F), that when set to true should output the information you need for your analysis to a file named towhee_chempotdata. Let me know how this works for you and if there is any information that you think should also be output to that file. thanks, Marcus
-
version 8.2.0
-
Thanks for reminding me of this bug. I found the issue and fixed it. I need to finish implementing one other feature and then I will release version 8.2.0 with this problem resolved. Marcus
-
On Mon, Mar 9, 2020 at 12:53 PM jonathan millis jmillis@users.sourceforge.net wrote: On 2020/03/04 06:44, Marcus Martin wrote: swapmoves.F is the main place where insertion Rosenbluth weights (often modified for the number density term) are stored. This happens using the twh_averages routine so that is a good keyword to search for in that subroutine and I believe the section starting at line 962 is the place to start. I suspect the real issue with the chemical potential averaging is that we in general...
-
The analyse_movie routine in the Utils file does not currently use preproc.h and instead reads the number of types of molecules directly from the towhee_movie file. Perhaps this was the case in an older version of Towhee (I no longer remember), but otherwise I am not sure how to reproduce the problem you were seeing. Marcus
-
swapmoves.F is the main place where insertion Rosenbluth weights (often modified for the number density term) are stored. This happens using the twh_averages routine so that is a good keyword to search for in that subroutine and I believe the section starting at line 962 is the place to start. I suspect the real issue with the chemical potential averaging is that we in general do a pretty mediocre job of sampling the chemical potential and I think there is still a lot of time spent on low Rosenbluth...
-
more minor bug fixes for version 8.1.2
-
version 8.1.2 preliminary
-
The first thing to keep in mind is that liquid pressures are quite noisy, espcially if your simulation is at the usual conditions of around 100 kPa of pressure. Given the settings I am guessing you are using for a single box NVT ensemble the only pressure being calculated uses the virial pressure method, as opposed to the thermodynamic pressure that comes from box size fluctuations. See the commentary on pressure for more information. http://towhee.sourceforge.net/algorithm/pressure.html You are...
-
Marcus Martin posted a comment on discussion Help
There is not a tabulated bond potential or angle potential, simply because there has not previously been any interest in such a thing. I would think most of the support for implementing such a potential already exists in various parts of Towhee since there are already spline functions for use with the tabulated embedded atom nonbonded potentials. In general, the bond and angle potentials are pretty easy to implement since they only appear in a couple of subroutines. Pretty much rwforcefield.F, vbond.F,...
-
Sorting through the files trying to figure out what is going wrong. Since I am not sure what the simulation is supposed to look like I need to ask some questions when I hit parts I think are strange to see if that is a problem, or a deliberate choice. The hmtraix values are unusual in your input file. hmatrix 37.2315 0 0 0.311405 65.2839 0 -6.37732 -12.9964 13.8888 Is this intentional? A "normal" brick-shaped box woud just have terms on the diagonals and the third line of this hmatrix is very unusual....
-
Seems most likely something is going wrong in the scripts that turn the input from LAMMPS into Towhee. Those rarely get tested and most still date back to when LAMMPS was written in FORTRAN. If you send me the relevant LAMMPS and Towhee files I can take a look to try and sort out what is happening. My first instinct is it probably is related to something going wrong with the coordinate frames for a molecule that wraps through the simulation box in complicated ways. If that is the case, then it should...
-
That is the fix I used in the now released 8.1.1. Not sure how this ever would have worked with the previous call, but given that it is for a special case of rigid three-member rings that are using arbitrary trial distributions for CBMC it probably just never came up. On the topic of NNBOND=6 this theoretically works now, but the way the arrays are staticly allocated setting NNBOND to this level creates a very large number of possible dihedral terms and I think this causes the memory allocation to...
-
Verified the error. Not sure how my compiler did not complain about this problem. Testing a fix now, expecting it to be incorporated into Towhee 8.1.1. Planning on releasing that version this evening. Marcus
-
I have several comments. 1) You really need to see on the order of hundreds of accepted molecule transfers in your simulation before the results are trustworthy. The example you provide only has 11 in a 100 cycle simulation. That is about the right number of transfers considering the number of cycles, but I would need to see something like a 10,000 cycle run output to comment on the answers. 2) I suggest using the rotational-bias molecule transfer move for this simulation as it is currently much...
-
Just in case anyone else ends up wondering about this issue the answer is since TraPPE-UA uses rigid bonds those bonds either have an energy of zero (within tolerance of the equilibrium bond length) or an extremely high energy (if not within tolerance of the equilibrium bond length). In almost any case where the conformation sampling algorithms are working properly, this results in zero energy vibration contributions. Marcus
-
Howdy, You have indeed found a typo, but it is in Table 2 of the TraPPE 2 paper, not in the implementation of TraPPE in Towhee. Please see for example the TraPPE-1 paper Equation 5 or look at the official TraPPE parameter website for n-alkanes (for example n-butane) and you will see the value of 355 is correct. Thank you for bringing this to my attention though. Never noticed that mistake in TraPPE-2 before and I am the person responsible for that error. I will add a note about it on the Towhee TraPPE-UA...
-
That should really not be happening so I am glad you reported this problem. Could you please also include the towhee_initial file you used for that run so I can reproduce and start tracking down where the virial goes nan? My initial guess is something related to using a fixed Ewald cutoff as that is the only thing you are doing that is different from my typical settings so perhaps there is a bug related to the updates from that feature. thanks, Marcus
-
I just released version 8.1.0 and that contains updates to the NERD family of forcefields and related bug fixes to make the Sum2003 cosine series work properly. Please give it a try and let me know if it works for your system as I have not fully tested the alkene parts yet. You will need to load both the NERD 2 and the Sum2003 files in towhee_input to make this work properly as now the Sum2003 file just contains the parameters in that paper (plus the ones that are implied in that paper, but are missing...
-
version 8.1.0
-
Most of that sounds good to me. However, I believe they deliberately changed their torsional parameter and that is the reason I broke the files into versions. If you check what I call the version 3 paper (Khare et al 2004 J. Phys. Chem. B 108, 10071-10076) you will see they again list that alkane dihedral potential and even though that paper decided to use kJ/mol instead of Kelvins, if you convert the numbers you find that torsion matches the Version 2 values and not the Version 1 values. I had not...
-
I am currently reimplementing the Sum 2003 potential and do not understand why those alkene groups were not included previously, and am working to remedy that situation. My current plan is to include all the parameters from the Sum2003 paper itself, and also from refereces 22-24 in that paper. I do not understand why I chose to implement the parameters from JCP 2001, 115, Nath and Khare in the previous version as that paper is not mentioned in Sum2003 so my current plan is to remove the quaternary...
-
I agree that the 1/2 is missing. Adding it into the ffsum2003.F file parameters and busy double checking everything in there for more typos. The V_0 parameter in their torsion sure seems pointless since it is always zero, but leaving that in place just in case there is some future work with a different value. Marcus
-
It looks like there is a significant bug in my implementation of this torsion. The equilibrium torsion and 4 parameters are listed in the ffsum2003.F file, but I believe the wrong number is set for ntorloop (3 instead of 4). This makes the output have one too few parameters, and therefore completely skips that final parameter when computing the torsional energies. In regards to the Sun2003 that looks like a typo that is in the user's manual and the code itself. Clearly this implementation has some...
-
Hard to answer that question from just a snippet of the input file. I can offer some suggestions based on my best guess of what might be going wrong. 1) I would use the default CBMC options unless you have a specific reason to not do so. I keep Towhee set so the defaults are what I believe to be the best options currently available in the code. The options you are using are a bit older and will hurt the acceptance rate compared to the defaults in most cases. 2) I would guess you are trying to simulate...
-
There is a logarithm involved when transforming the quantity that is averaged in Towhee into a chemical potential. This is why the block averages do not always nicely align with the total simulation average. Also, the chemical potential calculation is extremely noisy in the case where partial charges contribute significantly to the energy of the simulation. The general consensus is the configurational-bias insertion procedure for charged molecules is actively bad if you consider non-neutral groups...
-
I am not completely sure what you are asking. I get that you are finding the Sum2003 potential implemented in Towhee to lack parameters for alkene C=C double bonds, but that is because the paper does not clearly include those parameters. The only currently implemented support for double bonds involving Carbon is C=O. I see where it makes sense to include all of the parameters from Nerd V2 into this potential and I do not have any recollection why this was treated as a separate implementation instead...
-
Since Towhee has a one box identity switch move I am assuming the id switch move you implemented is a two box move. The volume move, particle swap move, and translation move have been used extensively and those results compared with many other workers so it is quite likely those are not the problem. That leaves the two most likely causes of disagreement as 1) The forcefields are implemented differently in your work versus the work of the Siepmann group and the Panagiatopoulos group. 2) The 2 box...
-
version 8.0.0 release commit
-
preliminary cleanup 8.0.0
-
preliminary 8.0.0
-
version 8.0.0 preliminary
-
That error message is almost always due to not having the full directory path listed in the towhee_parallel file. One common problem for new users is by default the Towhee examples assume everything is in a base directory named /towheebase and users will typically either need to symbolically link that location to their actual install location, or change the directory path in the towhee_parallel and towhee_input files (it is also used to find the forcefield input files inside of towhee_input). Attempting...
-
If you already have the desired coordinates for all of the H2O and CO2 molecules then this is possible using the towhee_coords file. If you are asking how to set up an initial lattice of just H2O and CO2 in the simulation box, but not in a specified region where you have placed a slab of another molecule then there is not currently a built-in method for accomplishing that task. You could hack something together by first setting up a simulation that is just the non-slab portion that contains your...
-
There is not an easy way to do this as Towhee has many different parameters stored in the restart files depending on the number of molecules in the simulation. If you are willing to edit the restart files by hand then it is possible to cut the positions of the molecules out of a 1 component simulation and paste them into a multicomponent restart file while leaving all of the other terms intact. It is not officially supported in the software (hence the cut and paste) but I have done it occasionally....
-
As you approach the critical temperature in a vapor-liquid coexistence curve the liquid and vapor branches are becoming more similar along with an increase in how much each phase fluctuates in density. This makes is more difficult to choose initial conditions such that the equilibrated simulation will result in a liquid and vapor box of the required minimum dimensions (generally twice the nonbonded cutoff). The answer to how to set up the simulation properly so that it will provide useful coexistence...
-
I have reproduced the problem and it looks like some of the data structures keeping track of backbone CBMC moves exceed the compiler limits for that large of a molecule. I'm afraid I lack a general answer to this problem as you are running into some of the memory limitations of the statically generated arrays in Towhee. Your best work around is to tune all of the MAX values in preproc.h to be as small as possible for your simulation and hope that makes the arrays small enough to compile. Unlikely...
-
I am confused about where you are having the problem. pdb2towhee just converts a pdb file into some files useful for input into Towhee. I do not understand how this interacts with using ssh. Often the pdb file needs to be cleaned up a bit to work properly with this utility, but the pdb is the only input that is required. Marcus
-
There is unfortunately not an example for the tabulated potential, so you have to...
-
The two most likely causes are a formatting issue with your input, or overlaps due...
-
version 7.2.0 final commit
-
7.2.0 bookkeeping commit
-
The beauty of Monte Carlo is just about any kind of move is theoretically possible....
-
interim 7.2.0 commit
-
That error most likely means there is something wrong with your compilers as there...
-
another primliminary 7.2.0 version
-
Not quite enough information there to allow me to say much about what is going wrong....
-
So pretty much the way to implement the "ghost volume change" move is along the lines...
-
version 7.2.0
-
The reason you can see these errors more with a 2 box simulation is you can get fluctuations...
-
There is not currently support for a cylindrical box in Towhee. I'm not entirely...
-
The new version 7.1.0 manual now has a discussion about the Polymer builder that...
-
version 7.1.0 final commit
-
I'm not sure if the code has really been tested much for RNA. I know Amalie Frischknecht...
-
preliminary 7.1.0, just forcfield files
-
Working my way through Towhee questions. Sorry for the very late reply. If you are...
-
Working through the Towhee questions. Sorry for the slow reply. There is absolutely...
-
1) There are always PBC in all three dimensions. 2) Yes, hard walls are often added...
-
Working my way through the backlog of Towhee questions. As of version 7.0.7 this...
-
All of the input file parameters are described in the Towhee online manual. Start...
-
Things are almost certainly out of date as the combination of Towhee and Tramonto...
-
Sorry for the late reply. Sounds like something is going wrong in the CBMC for this...
-
I expect the problem is with the assignment of improper torsions for your particular...
-
I am not completely understanding your question. If you have pdb structures for the...
-
Converting pdb files into useful input files almost always requires a fair bit of...
-
I discuss move probability allocation for vapor-liquid coexistence in section 4 of...
-
version 7.0.7
1 >