5. Basic Analysis

The MD Output

The output of the MD code itself (referred to as output in Section 2 is perhaps the most important simulation output to examine for two reasons:

It exists in plain text, so you can easily read it using vi, cat, tail, etc.
It provides a number of metrics that will quickly tell if your simulation has gone awry

Additionally, the output is regularly updated (thanks to iprint) which allows you to quickly tell if a simulation is going afoul without having to wait for the simulation to complete.

The output file generally shows the following information in the following order:

all of the inputted parameters from pressure.list, tape5, and all positions, velocities, and ltypes of all atoms in the starting configuration (either from knite12 or as generated by the MD code when ntype = 1).
the output of each iprint. This is generally the most useful data.
a summary of the data from the iprints (average and instantaneous temperature, energy, pressure)
positions, velocities, and ltypes of all atoms in the final configuration of the simulation

The data from each iprint contains the most immediately useful data; after opening the output file in vi, you can easily jump straight to this section by typing slash (/) to enter the text-search mode, then enter elaps and press return to jump straight to the first iprint's data. You will see output that should look something like this:

`elapsed time is 0.10000000E-11 ...`	This line begins each new print
`basic parameters nmol temp xl ... 1536 0.30000D+03 0.28525D-06 ...`	`nmol` is the number of atoms in your system `temp` is the temperature setting from tape5 `xl`, `yl`, and `zl` are the dimensions of your simulation box in cm `deltim` is the integration timestep from `tape`5 `rc` is the interaction cutoff distance from `tape`5
`reference value of total energy ...`	This is the value against which the system's energy is compared to calculate the energy drift (see below).
`average values for last 1000 ... potential energy -0.24248972E-10 ... temperature 0.32738174E+03 ... pressure 0.51779621E+05atm compressibility 0.23898885E-20`	These are the values for temperature, pressure, potential energy, and compressibility averaged since the last print. The quantities that enter into this average are sampled every `icheck` iterations.
`absolute value of energy drift ...`	The average energy drift in constant-energy simulations is an indicator of numerical error that enters into the simulation due to the limitations of the computer. This value would be zero on an infinitely precise computer; our simulations should have an E_drift < 10^-5 E_total
`instantaneous values potential energy -0.24256554E-10 kinetic energy 0.62109000E-13 total energy -0.24195585E-10 energy drift 0.11404493E-14 temperature 0.30000000E+03 three-body contribution to ...`	These are system properties from the configuration at which the iprint was initiated. During equilibration (`iequil` < `icalc`), the temperature here will always be equal to temp from tape5 even though the average temperature above may be different.
`restart data stored after 1000 iterations`	This line is printed every istore to indicate that the knite9 has been updated.

Note that the values printed are typically in CGS units; energy is in ergs (10^-7 J or 4.814 × 10⁺¹⁰ kcal) and dimensions are in cm.

Always examine the output file from your simulation after it is complete to verify that the simulation proceeded as you expected. At minimum, you should always verify that

for constant pressure runs, the system's dimensions and density haven't gone crazy. Check the first few prints and the last few prints.
for constant volume runs, the system's temperature hasn't shot up or frozen down. Again, check the first few and last few prints.
the simulation actually completed and didn't crash after a few thousand iterations. Search the output for the phrase "final positions."

Mistakes in tape5 can most often be caught by checking the above items, and virtually every student has lost weeks of progress at some point because those weeks of work were based upon a simulation that wasn't run correctly.

Next page: The Radial Distribution Function