Fork me on GitHub

Here are a list of frequently asked questions and their answers.

Q: What does DynamO stand for?

Short answer: Dynamics of Objects.

A more precise name would be "Dynamics of Discrete Objects," so if there is ever an O'Reilly book on DynamO it will feature a Dodo on the front cover.

Q: What is the logo of DynamO?

See descriptive caption below

The DynamO logo.

Short answer: An illustration of an electric dynamo (see left), which forms the O of DynamO.

The "source" of the drawing is the U.S. Patent 284,110 which describes an electric dynamo. The original figure was redrawn using inkscape. The artwork for the logo is available in the docs folder of the source code (available here).

Q: What are the units of DynamO?

Short answer: whatever units you use in the input configuration file.

The configuration files expose almost every parameter of the system (see below about $k_B$), so whatever units you use to setup the DynamO input, these are used as the units of the output. You must be consistent in your choice of units, so if the particle sizes are specified in Angstroms, your simulation size, particle positions, and all other parameters with units of length must also be specified in Angstroms. But, provided you are consistent in your choice of units, you may use any set of units you like.

There are two peculiarities which are discussed below:

Q: What units does the dynamod command use/produce?

Short answer: Whenever the dynamod command is used to generate a configuration, it tries to use the most natural set of dimensionless units for that type of system. The results should then directly correspond to the typical units used in publications.

For example, in a mono-component hard sphere system, the natural unit of length is the particle diameter, $\sigma$. The natural unit of mass is the mass of a single particle, $m$. The natural unit of time is more difficult to choose but we can pick a unit of energy and, through the definition of unit length and mass, use it to set the unit of time. The only property which has units of energy in the hard sphere system is the temperature $k_B\,T$. Therefore, the natural unit of time is $\sqrt{m\,\sigma^2 /k_B\,T}$! If we use these units of length ($\sigma=1$), mass ($m=1$), and time ($\sqrt{m\,\sigma^2 /k_B\,T}=1$) to reduce our system into dimensionless units, we find that the temperature, particle diameter, and particle mass always go to 1. In fact, this dimensional analysis of the system reveals that the only remaining independent parameters of the system are the number of particles, $N$, and the volume of the system, $V$. Typically, the volume of the system is expressed through the reduced number density $\rho^*=\sigma^3\,N/V$. Therefore, the only parameters ever varied in a study of monocomponent hard-sphere systems are the dimensionless density $\rho^*$ and the particle count.

The dynamod mode for creating hard sphere systems therefore sets the mass and diameter of particles to be equal to 1, as this is the natural dimensionless units of the system.

If we look at mono-component square-well systems we now have two properties with units of energy (and therefore time), the temperature $k_B\,T$ and the well-depth $\varepsilon$. It is common to select the well-depth as the unit of energy ($\varepsilon=1$), which results in an additional independent parameter $k_B\,T/\varepsilon$ on top of the parameters used for hard-spheres ($N$ and $\rho^*$). In this case, the unit of time is $\sqrt{m\,\sigma^2 /\varepsilon}$.

The dynamod mode for creating square-well systems therefore sets the mass, diameter, and well depth of all particles to be equal to 1, as this is the natural dimensionless units of the system.

Q: How does DynamO collect exact time-averages?

Short answer: Many properties have a constant value between events, and so the following equation for the time average can be solved in a piecewise analytical fashion: \[\left\langle A\right\rangle=t_{sim}^{-1}\int_0^{t_{sim}}A(t)\,{\rm d}t\] where $A(t)$ is the value of the property being averaged at a time $t$ and $t_{sim}$ is the duration of the simulation. To illustrate, if $N_{events}$ occur during the duration of the simulation, and the property $A$ only changes on events, the integration can be rewritten as a sum: \[\left\langle A\right\rangle=t_{sim}^{-1}\sum_{i=1}^{N_{events}}A(t_{i-1})\left(t_i-t_{i-1}\right)\] where $t_i$ is the time of the $i$th event, $t_0=0$ by definition, and $A(t_i)$ is the value of the property just after the $i$th event.

There are some restrictions to this. For example, in systems without gravity and without shearing boundary conditions, the velocities of all particles are constant between events. If we calculate the kinetic energy at the start of the simulation, the only changes will occur during the execution of an event. It is then very simple to collect exact time-averages using the equation above. However, in gravity and in shearing systems the particles velocities change with time causing the kinetic energy to vary between events. Some of these changes can also be integrated exactly, but at the moment DynamO does not do this.

Q: You say I can edit the configuration files using my favourite text editor? How?

Short answer: Most modern text editors bundled with Linux such as Kate or GEdit (my personal choice is emacs) support opening xml files and will even highlight them for you. Just open them using either the open file dialog or by calling the editor from the command line. For example, to open a compressed configuration file from the terminal to edit it using Kate, just write:

kate config.out.xml.bz2

DynamO writes out a compressed xml files by default (indicated by the extension .bz2). Luckily, most text editors will happily uncompress and recompress the file for you if you just open it directly, as shown above.

Q: When I try to run dynavis I get a GLXBadFBConfig error, what is wrong?

Short answer: Your graphics card and/or drivers do not support OpenGL 3.3 contexts, which are required for the visualisation library. To run the visualiser you will need a modern graphics card (your PC should be able to run modern games at decent speeds). You may need to either upgrade your drivers and/or your graphics card.

This error usually appears right at the start, and looks like:

dynarun  Copyright (C) 2011  Marcus N Campbell Bannerman
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under certain conditions. See the licence you obtained with
the code
Simulation: Reading the XML input file into memory
Simulation: Parsing the raw XML
Simulation: Loading tags from the XML
CellNeighbourList: Cells Loaded
NbListScheduler: Neighbour List Scheduler Algorithmn Loaded
Dynamics: Loading Particle Data
Dynamics: Particle count 1372
SystemInteraction: System halt set for 1.79769e+308
The program 'dynavis' received an X Window System error.
This probably reflects a bug in the program.
The error was 'GLXBadFBConfig'.
  (Details: serial 34 error_code 181 request_code 154 minor_code 34)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)
BoundedPQ: Exception Events = 0

The GLXBadFBConfig error is raised as your system was unable to provide an OpenGL 3.3 context.

If you are using built-in Intel graphics, you will probably not be able to use the visualiser yet until the open source drivers catch up with the closed-source drivers (and windows). They are catching up quickly but it will be 2014 or possibly later before OpenGL 3.3 is fully supported and distributed.

If you have a discrete graphics card, we highly recommend that you use the binary drivers. Binary drivers are the closed-source drivers developed by the manufacturers of your graphics card. For AMD/ATI cards, the packages are called FGLRX, but for NVidia cards they are usually called nvidia or nvidia-current.

Q: The Andersen thermostat is giving me a non-zero system momentum average. Is this an error?

Short answer: Probably not. The Andersen thermostat allows the system momentum to fluctuate about zero. This can lead to some suprisingly large values of the total momentum.

Please see issue #29 in the github tracker for more information.

Q: How do I stop/pause/output-data during a simulation?

Short answer: Press ctrl-c in the terminal you are running dynarun in, then press "s" and enter.

If dynarun recieves a SIGINT (which can be created in a terminal by pressing ctrl-c), the first time it recieves one it will try to present a menu to the user to allow them to either shutdown the simulation, peek at the data collected so far, or to look at any live statistics of the simulation. An example of the menu is presented below.

Tue 09:55, Events 100k, t 19.0976, <MFT> 0.131009, T 1, U 0
^C
Caught SIGINT, notifying running simulation...

<S>hutdown or <P>eek at data output:

While this menu is displayed, dynarun is paused and waiting for input. If dynarun recieves a second SIGINT (ctrl-c) while waiting for a response, it will immediately exit the program and no data is saved.

If you press "s" or "S" then press return or enter, dynamo will set the simulation to shutdown after the next event. This causes dynarun to exit normally, writing out the configuration file and any collected data as though it had naturally reached the end of a simulation.

If you press "p" or "P" followed by enter, dynamo will write out the collected data of the simulation to a file called peek.data.xml.bz2. This allows you to "peek" at the results collected so far without stopping the simulation.

Any other key followed by enter or return will cause dynamo to continue the simulation without any changes.

Q: Why do many simulations use packing fraction ($\eta$) instead of number density ($\rho$)?

Short answer: Volume fraction has reasonably well-established bounds whereas number densities depend strongly on the polydispersity of the system.

It is most convenient to work in packing fractions instead of densities as almost all systems have a maximum packing fraction somewhere near the mono-component hard sphere limit of $\eta^{max}_{HS}=\pi\,\sqrt{2}/6\approx0.74$, thus a system with a packing fraction near 0.6-0.7 is usually a high-density system regardless of the interactions. Number densities on the other hand have a varying range of values depending on the unit length scale and particle sizes.

For example, in the system studied in tutorial 4, once compression is complete, the packing fraction of $\eta=0.3$ has a reduced number density of $\rho\approx22$ whereas the mono-component hard sphere system has a maximum number density of $\rho_{HS}^{max}=\sqrt{2}\approx1.41$. Thus we cannot immediately estimate what a "high" number density is for a polydisperse system.

Q: How do I take "snapshots" or regular samples of a running simulation?

Short answer: Use the --snapshot or --snapshot-events options to dynarun.

The --snapshot option allows you to set a time interval after which dynarun will take a "snapshot" and save the configuration and output data of the system into a file named Snapshot.0.xml.bz2, Snapshot.1.xml.bz2, ... At the same time, the current data from the output plugins is saved to Snapshot.output.0.xml.bz2 etc.

If you use the --snapshot-events option, this allows you to take a snapshot every fixed number of events. These files are saved to Snapshot.output.Xe.xml.bz2 and Snapshot.output.Xe.xml.bz2 where X is replaced with the snapshot number. This name difference is to allow you to collect both event and time snapshots together.

Q: Why are my particle positions negative?

Short answer: DynamO can work with any sign of position, but for PBC simulations it assumes your simulation is centered on the coordinate $(0,0,0)$.

When DynamO writes out configuration files for simulations with Periodic Boundary Conditions (PBC), all particles are "wrapped" into the main image which is in the range $(\pm L_x/2,\, \pm L_y/2,\, \pm L_z/2)$ and $L_x$, $L_y$, and $L_z$ are given by the SimulationSize tag.

If the --unwrapped option is passed to dynarun, then all positions are written out in "unfolded" coordinates. These are the end positions of the particles that were originally in the primary image at the start of the simulation. This is handy when calculating diffusion properties (and is actually how DynamO stores the particle positions during the calculations).

Page last modified: Friday 28th November 2014