[lammps-users] Pair distance > table outer cutoff

vishnuvk · June 1, 2021, 12:34pm

Hi,

I've been running some simulations on small systems (4096 particles) that
involve `deform`s and `minimize`s. When I scale it up to a large system
(64,000 particles) and use the same script, I get the following error:

Pair distance > table outer cutoff

Why is that an error? If a pair distance is outside the cutoff as specified in
the 'table' file, why is LAMMPS attempting to calculate the interaction between
them?

Thank you!
Vishnu

akohlmey · June 1, 2021, 12:59pm

Hi,

I’ve been running some simulations on small systems (4096 particles) that
involve deforms and minimizes. When I scale it up to a large system
(64,000 particles) and use the same script, I get the following error:

Pair distance > table outer cutoff

Why is that an error? If a pair distance is outside the cutoff as specified in
the ‘table’ file, why is LAMMPS attempting to calculate the interaction between
them?

it doesn’t. it will only print this error if inconsistent data is provided in the input.

axel.

vishnuvk · June 1, 2021, 1:20pm

Hi Axel,

Yes, so I was hoping that the specific error message would narrow-down the
possible places where there might be inconsistencies. Does it tell you where
the issue likely is?

Thanks!
Vishnu

akohlmey · June 1, 2021, 1:40pm

No

vishnuvk · August 11, 2021, 12:12pm

The error points to the following line:

github.com

lammps/lammps/blob/c84e7a504087cc4e2132cf61bd513ab4bdc2bc09/src/pair_table.cpp#L138

    
      
          } else if (tabstyle == LINEAR) {
            itable = static_cast<int> ((rsq - tb->innersq) * tb->invdelta);
            if (itable >= tlm1)
              error->one(FLERR,"Pair distance > table outer cutoff: "
                                           "ijtype {} {} dist {}",itype,jtype,sqrt(rsq));
            fraction = (rsq - tb->rsq[itable]) * tb->invdelta;
            value = tb->f[itable] + fraction*tb->df[itable];
            fpair = factor_lj * value;
          } else if (tabstyle == SPLINE) {
            itable = static_cast<int> ((rsq - tb->innersq) * tb->invdelta);
            if (itable >= tlm1)
              error->one(FLERR,"Pair distance > table outer cutoff: "
                                           "ijtype {} {} dist {}",itype,jtype,sqrt(rsq));
            b = (rsq - tb->rsq[itable]) * tb->invdelta;
            a = 1.0 - b;
            value = a * tb->f[itable] + b * tb->f[itable+1] +
              ((a*a*a-a)*tb->f2[itable] + (b*b*b-b)*tb->f2[itable+1]) *
              tb->deltasq6;
            fpair = factor_lj * value;
          } else {
            rsq_lookup.f = rsq;

Does it occur anytime particles in the neighbour list are outside a cut-off
distance? That doesn't make sense, because that' is the point of a finite skin
distance.

The error does not occur, however, when I use the following setting:

neighbor 0.0 bin

The zero skin means that the neighbor-list ends up being created at nearly
every step (because the check occurs when anything moves more than half the
skin-distance), which does make the simulation run much longer.

This was necessary because the 'check' option is forced to 'yes' when
performing minimizations.

How might I fix this without incurring the expense of neighbor list
computations at every step?

Thank you!
Vishnu