lmp: support other units for both pair deepmd and fix dplr (#2800)

This PR is intended to do 2 things: 1. support other unit styles like si and nano (except lj); 2. support the non-metal units for fix dplr and compute deeptensor/atom. Unittests for both features mentioned above have also been added in this pull request. --------- Signed-off-by: Yifan Li李一帆 <[email protected]> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Jinzhe Zeng <[email protected]>
deepmodeling · Sep 18, 2023 · 19c6a68 · 19c6a68
1 parent 0d5737f
commit 19c6a68
Show file tree

Hide file tree

Showing 12 changed files with 406 additions and 176 deletions.
diff --git a/doc/third-party/lammps-command.md b/doc/third-party/lammps-command.md
@@ -1,5 +1,14 @@
 # LAMMPS commands
 
+## units
+All units in LAMMPS except `lj` are supported. `lj` is not supported.
+
+The most commonly used units are `metal`, since the internal units of distance, energy, force, and charge in DeePMD-kit are `\AA`, `eV`, `eV / \AA`, and `proton charge`, respectively. These units are consistent with the `metal` units in LAMMPS.
+
+If one wants to use other units like `real` or `si`, it is welcome to do so. There is no need to do the unit conversion mannualy. The unit conversion is done automatically by LAMMPS.
+
+The only thing that one needs to take care is the unit of the output of `compute deeptensor/atom`. Working with `metal` units for `compute deeptensor/atom` is totally fine, since there is no unit conversion. For other unit styles, we currently assume that the output of the `compute deeptensor/atom` command has the unit of distance and have applied the unit conversion factor of distance. If a user wants to infer quantities with units other than distance, the user is encouraged to open a GitHub feature request, so that the unit conversion factor can be added.
+
 ## Enable DeePMD-kit plugin (plugin mode)
 
 If you are using the plugin mode, enable DeePMD-kit package in LAMMPS with `plugin` command:
@@ -119,6 +128,7 @@ dump            1 all custom 100 water.dump id type c_dipole[1] c_dipole[2] c_di
 
 ### Restrictions
 - The `deeptensor/atom` compute is provided in the USER-DEEPMD package, which is compiled from the DeePMD-kit, visit the [DeePMD-kit website](https://github.com/deepmodeling/deepmd-kit) for more information.
+- For the issue of using a unit style for `compute deeptensor/atom`, refer to the discussions in [units](#units) of this page.
 
 
 ## Long-range interaction

diff --git a/source/lmp/compute_deeptensor_atom.cpp b/source/lmp/compute_deeptensor_atom.cpp
@@ -26,12 +26,12 @@ using namespace LAMMPS_NS;
 
 ComputeDeeptensorAtom::ComputeDeeptensorAtom(LAMMPS *lmp, int narg, char **arg)
     : Compute(lmp, narg, arg), dp(lmp), tensor(nullptr) {
-  if (!(strcmp(update->unit_style, "metal") == 0 ||
-        strcmp(update->unit_style, "real") == 0)) {
-    error->all(
-        FLERR,
-        "Compute deeptensor/atom requires metal or real unit; please set it by "
-        "\"units metal\" or \"units real\"");
+  if (strcmp(update->unit_style, "lj") == 0) {
+    error->all(FLERR,
+               "Compute deeptensor/atom does not support unit style lj. Please "
+               "use other "
+               "unit styles like metal or real unit instead. You may set it by "
+               "\"units metal\" or \"units real\"");
   }
 
   if (narg < 4) {
@@ -57,6 +57,8 @@ ComputeDeeptensorAtom::ComputeDeeptensorAtom(LAMMPS *lmp, int narg, char **arg)
   timeflag = 1;
 
   nmax = 0;
+
+  dist_unit_cvt_factor = force->angstrom;
 }
 
 /* ---------------------------------------------------------------------- */
@@ -116,16 +118,17 @@ void ComputeDeeptensorAtom::compute_peratom() {
     dtype[ii] = type[ii] - 1;
   }
   // get box
-  dbox[0] = domain->h[0];  // xx
-  dbox[4] = domain->h[1];  // yy
-  dbox[8] = domain->h[2];  // zz
-  dbox[7] = domain->h[3];  // zy
-  dbox[6] = domain->h[4];  // zx
-  dbox[3] = domain->h[5];  // yx
+  dbox[0] = domain->h[0] / dist_unit_cvt_factor;  // xx
+  dbox[4] = domain->h[1] / dist_unit_cvt_factor;  // yy
+  dbox[8] = domain->h[2] / dist_unit_cvt_factor;  // zz
+  dbox[7] = domain->h[3] / dist_unit_cvt_factor;  // zy
+  dbox[6] = domain->h[4] / dist_unit_cvt_factor;  // zx
+  dbox[3] = domain->h[5] / dist_unit_cvt_factor;  // yx
   // get coord
   for (int ii = 0; ii < nall; ++ii) {
     for (int dd = 0; dd < 3; ++dd) {
-      dcoord[ii * 3 + dd] = x[ii][dd] - domain->boxlo[dd];
+      dcoord[ii * 3 + dd] =
+          (x[ii][dd] - domain->boxlo[dd]) / dist_unit_cvt_factor;
     }
   }
 
@@ -155,7 +158,7 @@ void ComputeDeeptensorAtom::compute_peratom() {
     // record when selected and in group
     if (selected && ingroup) {
       for (int jj = 0; jj < size_peratom_cols; ++jj) {
-        tensor[ii][jj] = atensor[iter_tensor + jj];
+        tensor[ii][jj] = atensor[iter_tensor + jj] * dist_unit_cvt_factor;
       }
     }
     // if not selected or not in group set to 0.

diff --git a/source/lmp/compute_deeptensor_atom.h b/source/lmp/compute_deeptensor_atom.h
@@ -36,6 +36,7 @@ class ComputeDeeptensorAtom : public Compute {
   void compute_peratom() override;
   double memory_usage() override;
   void init_list(int, class NeighList *) override;
+  double dist_unit_cvt_factor;
 
  private:
   int nmax;

diff --git a/source/lmp/fix_dplr.cpp b/source/lmp/fix_dplr.cpp
@@ -61,10 +61,11 @@ FixDPLR::FixDPLR(LAMMPS *lmp, int narg, char **arg)
   qe2f = force->qe2f;
   xstyle = ystyle = zstyle = NONE;
 
-  if (strcmp(update->unit_style, "metal") != 0) {
-    error->all(
-        FLERR,
-        "Fix dplr requires metal unit, please set it by \"units metal\"");
+  if (strcmp(update->unit_style, "lj") == 0) {
+    error->all(FLERR,
+               "Fix dplr does not support unit style lj. Please use other "
+               "unit styles like metal or real unit instead. You may set it by "
+               "\"units metal\" or \"units real\"");
   }
 
   int iarg = 3;
@@ -142,6 +143,9 @@ FixDPLR::FixDPLR(LAMMPS *lmp, int narg, char **arg)
   if (!pair_deepmd) {
     error->all(FLERR, "pair_style deepmd should be set before this fix\n");
   }
+  ener_unit_cvt_factor = pair_deepmd->ener_unit_cvt_factor;
+  dist_unit_cvt_factor = pair_deepmd->dist_unit_cvt_factor;
+  force_unit_cvt_factor = pair_deepmd->force_unit_cvt_factor;
 
   int n = atom->ntypes;
   std::vector<std::string> type_names = pair_deepmd->type_names;
@@ -445,16 +449,17 @@ void FixDPLR::pre_force(int vflag) {
     dtype[ii] = type_idx_map[type[ii] - 1];
   }
   // get box
-  dbox[0] = domain->h[0];  // xx
-  dbox[4] = domain->h[1];  // yy
-  dbox[8] = domain->h[2];  // zz
-  dbox[7] = domain->h[3];  // zy
-  dbox[6] = domain->h[4];  // zx
-  dbox[3] = domain->h[5];  // yx
+  dbox[0] = domain->h[0] / dist_unit_cvt_factor;  // xx
+  dbox[4] = domain->h[1] / dist_unit_cvt_factor;  // yy
+  dbox[8] = domain->h[2] / dist_unit_cvt_factor;  // zz
+  dbox[7] = domain->h[3] / dist_unit_cvt_factor;  // zy
+  dbox[6] = domain->h[4] / dist_unit_cvt_factor;  // zx
+  dbox[3] = domain->h[5] / dist_unit_cvt_factor;  // yx
   // get coord
   for (int ii = 0; ii < nall; ++ii) {
     for (int dd = 0; dd < 3; ++dd) {
-      dcoord[ii * 3 + dd] = x[ii][dd] - domain->boxlo[dd];
+      dcoord[ii * 3 + dd] =
+          (x[ii][dd] - domain->boxlo[dd]) / dist_unit_cvt_factor;
     }
   }
   // get lammps nlist
@@ -523,9 +528,11 @@ void FixDPLR::pre_force(int vflag) {
     int res_idx = sel_fwd[idx0];
     // int ret_idx = dpl_bwd[res_idx];
     for (int dd = 0; dd < 3; ++dd) {
-      x[idx1][dd] = x[idx0][dd] + tensor[res_idx * 3 + dd];
+      x[idx1][dd] =
+          x[idx0][dd] + tensor[res_idx * 3 + dd] * dist_unit_cvt_factor;
       // res_buff[idx1 * odim + dd] = tensor[res_idx * odim + dd];
-      dipole_recd[idx0 * 3 + dd] = tensor[res_idx * 3 + dd];
+      dipole_recd[idx0 * 3 + dd] =
+          tensor[res_idx * 3 + dd] * dist_unit_cvt_factor;
     }
   }
   // cout << "-------------------- fix/dplr: pre force " << endl;
@@ -568,17 +575,18 @@ void FixDPLR::post_force(int vflag) {
     for (int ii = 0; ii < nall; ++ii) {
       dtype[ii] = type_idx_map[type[ii] - 1];
     }
-    dbox[0] = domain->h[0];  // xx
-    dbox[4] = domain->h[1];  // yy
-    dbox[8] = domain->h[2];  // zz
-    dbox[7] = domain->h[3];  // zy
-    dbox[6] = domain->h[4];  // zx
-    dbox[3] = domain->h[5];  // yx
+    dbox[0] = domain->h[0] / dist_unit_cvt_factor;  // xx
+    dbox[4] = domain->h[1] / dist_unit_cvt_factor;  // yy
+    dbox[8] = domain->h[2] / dist_unit_cvt_factor;  // zz
+    dbox[7] = domain->h[3] / dist_unit_cvt_factor;  // zy
+    dbox[6] = domain->h[4] / dist_unit_cvt_factor;  // zx
+    dbox[3] = domain->h[5] / dist_unit_cvt_factor;  // yx
     // get coord
     double **x = atom->x;
     for (int ii = 0; ii < nall; ++ii) {
       for (int dd = 0; dd < 3; ++dd) {
-        dcoord[ii * 3 + dd] = x[ii][dd] - domain->boxlo[dd];
+        dcoord[ii * 3 + dd] =
+            (x[ii][dd] - domain->boxlo[dd]) / dist_unit_cvt_factor;
       }
     }
     // revise force according to efield
@@ -599,7 +607,7 @@ void FixDPLR::post_force(int vflag) {
     for (int ii = 0; ii < nlocal; ++ii) {
       double tmpf[3];
       for (int dd = 0; dd < 3; ++dd) {
-        tmpf[dd] = q[ii] * efield[dd];
+        tmpf[dd] = q[ii] * efield[dd] * force->qe2f;
       }
       for (int dd = 0; dd < 3; ++dd) {
         dfele[ii * 3 + dd] += tmpf[dd];
@@ -632,8 +640,17 @@ void FixDPLR::post_force(int vflag) {
   vector<FLOAT_PREC> dfcorr, dvcorr;
   // compute
   try {
+    for (int ii = 0; ii < nlocal * 3; ++ii) {
+      dfele[ii] /= force_unit_cvt_factor;
+    }
     dtm.compute(dfcorr, dvcorr, dcoord, dtype, dbox, valid_pairs, dfele, nghost,
                 lmp_list);
+    for (int ii = 0; ii < nlocal * 3; ++ii) {
+      dfcorr[ii] *= force_unit_cvt_factor;
+    }
+    for (int ii = 0; ii < 9; ++ii) {
+      dvcorr[ii] *= ener_unit_cvt_factor;
+    }
   } catch (deepmd_compat::deepmd_exception &e) {
     error->one(FLERR, e.what());
   }

diff --git a/source/lmp/fix_dplr.h b/source/lmp/fix_dplr.h
@@ -54,6 +54,7 @@ class FixDPLR : public Fix {
   void unpack_reverse_comm(int, int *, double *) override;
   double compute_scalar(void) override;
   double compute_vector(int) override;
+  double ener_unit_cvt_factor, dist_unit_cvt_factor, force_unit_cvt_factor;
 
  private:
   PairDeepMD *pair_deepmd;