Dear lammps users,

I have recently started using the fix adapt and compute ti commands in

lammps, and had some queries.

Here is my input file. It is intended to tether particles of type 1 to

type 2 using the gauss pair interaction. The gaussian energy is

"switched on" slowly from 0 to 1. Type 1 particles interact via a

truncated LJ potential.

As the first step, I would like to compute the dU/dl for l=0

U_s=-lambda U_g

dU_s/dl = -U_g

f(lambda)=-lambda

df/dl = -1.0

Unfortunately I get all zeros from the compute ti.

Is this an expected behavior ?

According to the relations above, the value of the derivative would be

the value of the original gauss potential computed for the ensemble

generated.

The best person to answer this is Sai (CCd).

He's in the middle of a vacation and moving to

a new position at MIT, so not sure how quickly

he will see this.

Steve

The derivative at lambda=0 return all 0's to avoid undefined calculations.

dU(lambda)/dlambda = U_unscaled*f'(lambda)/f(lambda)

will be undefined when f(lambda)=0. Use a small f(lambda) (0.01 or smaller) to get around this issue.

Sai.

Thank you, Sai. I had indeed tried small values - and they work well.

Perhaps a comment/note in the compute ti manual page might be added -

as a pointer to beginners? Functions like (1-lambda) will also give

zeros at the boundary...

Best,

Manish Agarwal

<[email protected]>

Postdoctoral Research Scientist

Chemical Engineering

Columbia University

- - - - - - - - - - - - - - - - - - - - - - - - - - -