pair_style snap command

Accelerator Variants: snap/intel, snap/kk


pair_style snap


pair_style snap
pair_coeff * * InP.snapcoeff InP.snapparam In In P P


Pair style snap defines the spectral neighbor analysis potential (SNAP), a machine-learning interatomic potential (Thompson). Like the GAP framework of Bartok et al. (Bartok2010), SNAP uses bispectrum components to characterize the local neighborhood of each atom in a very general way. The mathematical definition of the bispectrum calculation and its derivatives w.r.t. atom positions is identical to that used by compute snap, which is used to fit SNAP potentials to ab initio energy, force, and stress data. In SNAP, the total energy is decomposed into a sum over atom energies. The energy of atom i is expressed as a weighted sum over bispectrum components.

\[E^i_{SNAP}(B_1^i,...,B_K^i) = \beta^{\mu_i}_0 + \sum_{k=1}^K \beta_k^{\mu_i} B_k^i\]

where \(B_k^i\) is the k-th bispectrum component of atom i, and \(\beta_k^{\mu_i}\) is the corresponding linear coefficient that depends on \(\mu_i\), the SNAP element of atom i. The number of bispectrum components used and their definitions depend on the value of twojmax and other parameters defined in the SNAP parameter file described below. The bispectrum calculation is described in more detail in compute sna/atom.

Note that unlike for other potentials, cutoffs for SNAP potentials are not set in the pair_style or pair_coeff command; they are specified in the SNAP potential files themselves.

Only a single pair_coeff command is used with the snap style which specifies a SNAP coefficient file followed by a SNAP parameter file and then N additional arguments specifying the mapping of SNAP elements to LAMMPS atom types, where N is the number of LAMMPS atom types:

  • SNAP coefficient file

  • SNAP parameter file

  • N element names = mapping of SNAP elements to atom types

As an example, if a LAMMPS indium phosphide simulation has 4 atoms types, with the first two being indium and the third and fourth being phophorous, the pair_coeff command would look like this:

pair_coeff * * snap InP.snapcoeff InP.snapparam In In P P

The first 2 arguments must be * * so as to span all LAMMPS atom types. The two filenames are for the coefficient and parameter files, respectively. The two trailing ‘In’ arguments map LAMMPS atom types 1 and 2 to the SNAP ‘In’ element. The two trailing ‘P’ arguments map LAMMPS atom types 3 and 4 to the SNAP ‘P’ element.

If a SNAP mapping value is specified as NULL, the mapping is not performed. This can be used when a snap potential is used as part of the hybrid pair style. The NULL values are placeholders for atom types that will be used with other potentials.

The name of the SNAP coefficient file usually ends in the “.snapcoeff” extension. It may contain coefficients for many SNAP elements. The only requirement is that each of the unique element names appearing in the LAMMPS pair_coeff command appear exactly once in the SNAP coefficient file. It is okay if the SNAP coefficient file contains additional elements not in the pair_coeff command, except when using chemflag (see below). The name of the SNAP parameter file usually ends in the “.snapparam” extension. It contains a small number of parameters that define the overall form of the SNAP potential. See the pair_coeff page for alternate ways to specify the path for these files.

SNAP potentials are quite commonly combined with one or more other LAMMPS pair styles using the hybrid/overlay pair style. As an example, the SNAP tantalum potential provided in the LAMMPS potentials directory combines the snap and zbl pair styles. It is invoked by the following commands:

variable zblcutinner equal 4
variable zblcutouter equal 4.8
variable zblz equal 73
pair_style hybrid/overlay &
zbl ${zblcutinner} ${zblcutouter} snap
pair_coeff * * zbl 0.0
pair_coeff 1 1 zbl ${zblz}
pair_coeff * * snap Ta06A.snapcoeff Ta06A.snapparam Ta

It is convenient to keep these commands in a separate file that can be inserted in any LAMMPS input script using the include command.

The top of the SNAP coefficient file can contain any number of blank and comment lines (start with #), but follows a strict format after that. The first non-blank non-comment line must contain two integers:

  • nelem = Number of elements

  • ncoeff = Number of coefficients

This is followed by one block for each of the nelem elements. The first line of each block contains three entries:

  • Element name (text string)

  • R = Element radius (distance units)

  • w = Element weight (dimensionless)

This line is followed by ncoeff coefficients, one per line.

The SNAP parameter file can contain blank and comment lines (start with #) anywhere. Each non-blank non-comment line must contain one keyword/value pair. The required keywords are rcutfac and twojmax. Optional keywords are rfac0, rmin0, switchflag, bzeroflag, quadraticflag, chemflag, bnormflag, wselfallflag, switchinnerflag, sinner, dinner, chunksize, and parallelthresh.

The default values for these keywords are

  • rfac0 = 0.99363

  • rmin0 = 0.0

  • switchflag = 1

  • bzeroflag = 1

  • quadraticflag = 0

  • chemflag = 0

  • bnormflag = 0

  • wselfallflag = 0

  • switchinnerflag = 0

  • chunksize = 32768

  • parallelthresh = 8192

For detailed definitions of all of these keywords, see the compute sna/atom doc page.

If quadraticflag is set to 1, then the SNAP energy expression includes additional quadratic terms that have been shown to increase the overall accuracy of the potential without much increase in computational cost (Wood).

\[E^i_{SNAP}(\mathbf{B}^i) = \beta^{\mu_i}_0 + \boldsymbol{\beta}^{\mu_i} \cdot \mathbf{B}_i + \frac{1}{2}\mathbf{B}^t_i \cdot \boldsymbol{\alpha}^{\mu_i} \cdot \mathbf{B}_i\]

where \(\mathbf{B}_i\) is the K-vector of bispectrum components, \(\boldsymbol{\beta}^{\mu_i}\) is the K-vector of linear coefficients for element \(\mu_i\), and \(\boldsymbol{\alpha}^{\mu_i}\) is the symmetric K by K matrix of quadratic coefficients. The SNAP coefficient file should contain K(K+1)/2 additional coefficients in each element block, the upper-triangular elements of \(\boldsymbol{\alpha}^{\mu_i}\).

If chemflag is set to 1, then the energy expression is written in terms of explicit multi-element bispectrum components indexed on ordered triplets of elements, which has been shown to increase the ability of the SNAP potential to capture energy differences in chemically complex systems, at the expense of a significant increase in computational cost (Cusentino).

\[E^i_{SNAP}(\mathbf{B}^i) = \beta^{\mu_i}_0 + \sum_{\kappa,\lambda,\mu} \boldsymbol{\beta}^{\kappa\lambda\mu}_{\mu_i} \cdot \mathbf{B}^{\kappa\lambda\mu}_i\]

where \(\mathbf{B}^{\kappa\lambda\mu}_i\) is the K-vector of bispectrum components for neighbors of elements \(\kappa\), \(\lambda\), and \(\mu\) and \(\boldsymbol{\beta}^{\kappa\lambda\mu}_{\mu_i}\) is the corresponding K-vector of linear coefficients for element \(\mu_i\). The SNAP coefficient file should contain a total of \(K N_{elem}^3\) coefficients in each element block, where \(N_{elem}\) is the number of elements in the SNAP coefficient file, which must equal the number of unique elements appearing in the LAMMPS pair_coeff command, to avoid ambiguity in the number of coefficients.

The keyword switchinnerflag activates an additional switching function that smoothly turns off contributions to the SNAP potential from neighbor atoms at short separations. If switchinnerflag is set to 1 then the additional keywords sinner and dinner must also be provided. Each of these is followed by nelements values, where nelements is the number of unique elements appearing in appearing in the LAMMPS pair_coeff command. The element order should correspond to the order in which elements first appear in the pair_coeff command reading from left to right.

The keywords chunksize and parallelthresh are only applicable when using the pair style snap with the KOKKOS package on GPUs and are ignored otherwise. The chunksize keyword controls the number of atoms in each pass used to compute the bispectrum components and is used to avoid running out of memory. For example if there are 8192 atoms in the simulation and the chunksize is set to 4096, the bispectrum calculation will be broken up into two passes (running on a single GPU). The parallelthresh keyword controls a crossover threshold for performing extra parallelism. For small systems, exposing additional parallelism can be beneficial when there is not enough work to fully saturate the GPU threads otherwise. However, the extra parallelism also leads to more divergence and can hurt performance when the system is already large enough to saturate the GPU threads. Extra parallelism will be performed if the chunksize (or total number of atoms per GPU) is smaller than parallelthresh.


The previously used diagonalstyle keyword was removed in 2019, since all known SNAP potentials use the default value of 3.

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

For atom type pairs I,J and I != J, where types I and J correspond to two different element types, mixing is performed by LAMMPS with user-specifiable parameters as described above. You never need to specify a pair_coeff command with I != J arguments for this style.

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

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

This pair style can only be used via the pair keyword of the run_style respa command. It does not support the inner, middle, outer keywords.

Styles with a gpu, intel, kk, omp, or opt suffix are functionally the same as the corresponding style without the suffix. They have been optimized to run faster, depending on your available hardware, as discussed on the Accelerator packages page. The accelerated styles take the same arguments and should produce the same results, except for round-off and precision issues.

These accelerated styles are part of the GPU, INTEL, KOKKOS, OPENMP, and OPT packages, respectively. They are only enabled if LAMMPS was built with those packages. See the Build package page for more info.

You can specify the accelerated styles explicitly in your input script by including their suffix, or you can use the -suffix command-line switch when you invoke LAMMPS, or you can use the suffix command in your input script.

See the Accelerator packages page for more instructions on how to use the accelerated styles effectively.


This style is part of the ML-SNAP package. It is only enabled if LAMMPS was built with that package. See the Build package page for more info.

The snap/intel accelerator variant will only be available if LAMMPS is built with Intel compilers and for CPUs with AVX-512 support. While the INTEL package in general allows multiple floating point precision modes to be selected, snap/intel will currently always use full double precision regardless of the precision mode selected. Additionally, the intel variant of snap will NOT use multiple threads with OpenMP.



(Thompson) Thompson, Swiler, Trott, Foiles, Tucker, J Comp Phys, 285, 316 (2015).

(Bartok2010) Bartok, Payne, Kondor, Csanyi, Phys Rev Lett, 104, 136403 (2010).

(Wood) Wood and Thompson, J Chem Phys, 148, 241721, (2018)

(Cusentino) Cusentino, Wood, and Thompson, J Phys Chem A, xxx, xxxxx, (2020)