What erate means in LAMMPS

The definition

Engineering strain,per unit time

How fast the box changes length relative to its original length. Every confusion about it comes from forgetting that last word.

Relative to the original

Engineering strain divides by the starting length, not the current one. True strain uses the current length.

Per time, not per step

The rate is in inverse time units of your unit system, so the timestep changes how much strain a run accumulates.

It deforms the box

The cell changes shape and the atoms follow. Nothing is pulled directly — the boundary does the work.

Using it

One linein fix deform

erate appears as a keyword on fix deform, alongside the dimension you are deforming and what to do with the atoms inside.

in.deform

fix 1 all nve
fix 2 all deform 1 xy erate 0.001 units box remap x

thermo_style custom step temp pxy pe lx ly lz
thermo 100
dump 1 all custom 1000 shear.dump id type x y z
run 100000

deform 1Apply the deformation every timestep. A larger interval deforms in visible jumps rather than smoothly.

xy erate 0.001Shear the xy tilt at that engineering rate. Use x, y or z instead for tension and compression.

units boxInterpret the rate against box dimensions rather than lattice spacing, which is what you usually want.

remap xMove atom coordinates with the box. Without it, atoms stay put and the box slides past them.

The trap

Too fast,and you measure the rate

This is the single most common failure, and it does not announce itself — the run completes and the numbers look like data.

An MD simulation covers nanoseconds. A tensile test in a laboratory takes minutes. Closing that gap by pulling harder means straining at rates many orders of magnitude above anything experimental, and materials genuinely behave differently up there: yield stresses come out high, dislocation mechanisms change, and amorphous systems have no time to relax at all.

The fix is not a better value but a convergence check. Run the same setup at your rate and at half of it. If the stress-strain curve moves, you are measuring your strain rate rather than your material, and the only remedy is a slower pull — which is to say, more machine hours.

Start slow

Begin below the rate you think you need. It is easier to justify a slow run than to explain a fast one.

Run a zero control

erate 0 should produce a flat curve. If it does not, the problem is upstream of the deformation.

Match your manual version

fix deform keywords have changed across releases. Read the page for the build you are actually running.

Reading the output

Stress against strain,and then a picture

The thermo output gives you the curve. The dump file tells you what the material was doing while the curve was being drawn.

pxyPlotagainst accumulated strain

The dumpThen openand see where it localised

Halve itConfirmthe curve should not move

Print the relevant pressure-tensor component with thermo_style custom — pxy for a shear in xy — and plot it against accumulated strain. Then open the dump in OVITO or ParaView and look at where the deformation actually localised. A curve with a clean yield point and a snapshot showing the whole box shearing uniformly are telling you two different stories, and the snapshot is the honest one.

The status

LAMMPS isbeing built

The image and its definition are being verified. When they are done, it deploys by name like the rest.

Ask for itSay so on the community tracker, and it moves up the list.Or build itOpen a pull request on the hub, following a published one.

Read it, then run itPrepaid and by the hour, on a machine that is yours about a minute after you ask.

Start a machine All notes