pair_style local/density command


pair_style style arg
  • style = local/density

  • arg = name of file containing tabulated values of local density and the potential


pair_style local/density benzene_water.localdensity.table

pair_style hybrid/overlay table spline 500 local/density
pair_coeff * * local/density  benzene_water.localdensity.table


The local density (LD) potential is a mean-field manybody potential, and, in some way, a generalization of embedded atom models (EAM). The name “local density potential” arises from the fact that it assigns an energy to an atom depending on the number of neighboring atoms of a given type around it within a predefined spherical volume (i.e., within the cutoff). The bottom-up coarse-graining (CG) literature suggests that such potentials can be widely useful in capturing effective multibody forces in a computationally efficient manner and thus improve the quality of CG models of implicit solvation (Sanyal1) and phase-segregation in liquid mixtures (Sanyal2), and provide guidelines to determine the extent of manybody correlations present in a CG model (Rosenberger). The LD potential in LAMMPS is primarily intended to be used as a corrective potential over traditional pair potentials in bottom-up CG models via hybrid/overlay pair style with other explicit pair interaction terms (e.g., tabulated, Lennard-Jones, Morse etc.). Because the LD potential is not a pair potential per se, it is implemented simply as a single auxiliary file with all specifications that will be read upon initialization.


Thus when used as the only interaction in the system, there is no corresponding pair_coeff command and when used with other pair styles using the hybrid/overlay option, the corresponding pair_coeff command must be supplied * * as placeholders for the atom types.

System with a single CG atom type:

A system of a single atom type (e.g., LJ argon) with a single local density (LD) potential would have an energy given by:

\[U_{LD} = \sum_i F(\rho_i)\]

where \(\rho_i\) is the LD at atom i and \(F(\rho)\) is similar in spirit to the embedding function used in EAM potentials. The LD at atom i is given by the sum

\[\rho_i = \sum_{j \neq i} \varphi(r_{ij})\]

where \(\varphi\) is an indicator function that is one at r=0 and zero beyond a cutoff distance R2. The choice of the functional form of \(\varphi\) is somewhat arbitrary, but the following piecewise cubic function has proven sufficiently general: (Sanyal1), (Sanyal2) (Rosenberger)

\[\begin{split}\varphi(r) = \begin{cases} 1 & r \le R_1 \\ c_0 + c_2r^2 + c_4r^4 + c_6r^6 & r \in (R_1, R_2) \\ 0 & r \ge R_2 \end{cases}\end{split}\]

The constants c are chosen so that the indicator function smoothly interpolates between 1 and 0 between the distances R1 and R2, which are called the inner and outer cutoffs, respectively. Thus phi satisfies phi(R1) = 1, phi(R2) = dphi/dr @ (r=R1) = dphi/dr @ (r=R2) = 0. The embedding function F(rho) may or may not have a closed-form expression. To maintain generality, it is practically represented with a spline-interpolated table over a predetermined range of rho. Outside of that range it simply adopts zero values at the endpoints.

It can be shown that the total force between two atoms due to the LD potential takes the form of a pair force, which motivates its designation as a LAMMPS pair style. Please see (Sanyal1) for details of the derivation.

Systems with arbitrary numbers of atom types:

The potential is easily generalized to systems involving multiple atom types:

\[U_{LD} = \sum_i a_\alpha F(\rho_i)\]

with the LD expressed as

\[\rho_i = \sum_{j \neq i} b_\beta \varphi(r_{ij})\]

where \(\alpha\) gives the type of atom i, \(\beta\) the type of atom j, and the coefficients a and b filter for atom types as specified by the user. a is called the central atom filter as it determines to which atoms the potential applies; \(a_{\alpha} = 1\) if the LD potential applies to atom type \(\alpha\) else zero. On the other hand, b is called the neighbor atom filter because it specifies which atom types to use in the calculation of the LD; \(b_{\beta} = 1\) if atom type \(\beta\) contributes to the LD and zero otherwise.


Note that the potentials need not be symmetric with respect to atom types, which is the reason for two distinct sets of coefficients a and b. An atom type may contribute to the LD but not the potential, or to the potential but not the LD. Such decisions are made by the user and should (ideally) be motivated on physical grounds for the problem at hand.

General form for implementation in LAMMPS:

Of course, a system with many atom types may have many different possible LD potentials, each with their own atom type filters, cutoffs, and embedding functions. The most general form of this potential as implemented in the pair_style local/density is:

\[U_{LD} = \sum_k U_{LD}^{(k)} = \sum_i \left[ \sum_k a_\alpha^{(k)} F^{(k)} \left(\rho_i^{(k)}\right) \right]\]

where, k is an index that spans the (arbitrary) number of applied LD potentials N_LD. Each LD is calculated as before with:

\[\rho_i^{(k)} = \sum_j b_\beta^{(k)} \varphi^{(k)} (r_{ij})\]

The superscript on the indicator function phi simply indicates that it is associated with specific values of the cutoff distances R1(k) and R2(k). In summary, there may be N_LD distinct LD potentials. With each potential type (k), one must specify:

  • the inner and outer cutoffs as R1 and R2

  • the central type filter a(k), where k = 1,2,…N_LD

  • the neighbor type filter b(k), where k = 1,2,…N_LD

  • the LD potential function F(k)(rho), typically as a table that is later spline-interpolated

Tabulated input file format:

Line 1:             comment or blank (ignored)
Line 2:             comment or blank (ignored)
Line 3:             N_LD N_rho (# of LD potentials and # of tabulated values, single space separated)
Line 4:             blank (ignored)
Line 5:             R1(k) R2(k) (lower and upper cutoffs, single space separated)
Line 6:             central-types (central atom types, single space separated)
Line 7:             neighbor-types (neighbor atom types single space separated)
Line 8:             rho_min rho_max drho (min, max and diff. in tabulated rho values, single space separated)
Line 9:             F(k)(rho_min + 0.drho)
Line 10:            F(k)(rho_min + 1.drho)
Line 11:            F(k)(rho_min + 2.drho)
Line 9+N_rho:       F(k)(rho_min + N_rho . drho)
Line 10+N_rho:      blank (ignored)

Block 2

Block 3

Block N_LD

Lines 5 to 9+N_rho constitute the first block. Thus the input file is separated (by blank lines) into N_LD blocks each representing a separate LD potential and each specifying its own upper and lower cutoffs, central and neighbor atoms, and potential. In general, blank lines anywhere are ignored.

Mixing, shift, table, tail correction, restart, rRESPA info

This pair style does not support automatic mixing. For atom type pairs \(\alpha\), \(\beta\) and \(\alpha\) != \(\beta\), even if LD potentials of type (\(\alpha\), \(\alpha\)) and (\(\beta\), \(\beta\)) are provided, you will need to explicitly provide LD potential types (\(\alpha\), \(\beta\)) and (\(\beta\), \(\alpha\)) if need be (Here, the notation (\(\alpha\), \(\beta\)) means that \(\alpha\) is the central atom to which the LD potential is applied and \(\beta\) is the neighbor atom which contributes to the LD potential on \(\alpha\)).

This pair style does not support the pair_modify shift, table, and tail options.

The local/density pair style does not write its information to binary restart files, since it is stored in tabulated potential files. Thus, you need to re-specify the pair_style and pair_coeff commands in an input script that reads a restart file.


The local/density pair style is a part of the MANYBODY package. It is only enabled if LAMMPS was built with that package. See the Build package page for more info.



(Sanyal1) Sanyal and Shell, Journal of Chemical Physics, 2016, 145 (3), 034109.

(Sanyal2) Sanyal and Shell, Journal of Physical Chemistry B, 122 (21), 5678-5693.

(Rosenberger) Rosenberger, Sanyal, Shell and van der Vegt, Journal of Chemical Physics, 2019, 151 (4), 044111.