Loading doc/src/min_modify.txt +35 −28 Changes for doc/src/min_modify.txt: 35 added lines, 28 removed lines. Original line number Diff line number Diff line Loading @@ -19,19 +19,24 @@ keyword = {dmax} or {line} or {delaystep} or {dt_grow} or {dt_shrink} max = maximum distance for line search to move (distance units) {line} value = {backtrack} or {quadratic} or {forcezero} backtrack,quadratic,forcezero = style of linesearch to use {delaystep} value = {int} for adaptglok {dt_grow} value = {float} for adaptglok {dt_shrink} value = {float} for adaptglok {alpha0} value = {float} for adaptglok {alpha_shrink} value = {float} for adaptglok {tmax} value = {float} for adaptglok {tmin} value = {float} for adaptglok {delaystep} value = steps steps = number of steps for dynamics temporization, adaptglok only {dtgrow} value = grow grow = factor increasing the timestep, adaptglok only {dtshrink} value = shrink shrink = factor decreasing the timestep, adaptglok only {alpha0} value = alpha alpha = coefficient mixing velocities and forces, adaptglok only {alphashrink} value = shrink shrink = factor decreasing alpha, adaptglok only {tmax} value = coef coef = factor defining the maximum timestep, adaptglok only {tmin} value = coef coef = factor defining the minimum timestep, adaptglok only {integrator} value = {eulerimplicit} or {eulerexplicit} or {verlet} or {leapfrog} for adaptglok {halfstepback} value = {yes} or {no} for adaptglok :pre {relaxbox} value = {no} or {iso} or {axial} for adaptglok {relaxbox_modulus} value = {float} for adaptglok {relaxbox_rate} value = {float} for adaptglok or {leapfrog} eulerimplicit,eulerexplicit,verlet,leapfrog = integration scheme, adaptglok {halfstepback} value = {yes} or {no}, adaptglok only :pre :ule [Examples:] Loading Loading @@ -80,18 +85,20 @@ that difference may be smaller than machine epsilon even if atoms could move in the gradient direction to reduce forces further. The style {adaptglok} has several parameters that can be tuned in order to optimize the relaxation: {integrator}, {delaystep}, {dt_grow}, {dt_shrink}, {alpha0}, {alpha_shrink}, {tmax} and {tmin}. The default values have been chosen to be reliable in most cases, but one could be willing to adapt them for particular cases. In particular for debugging purpose, it is possible to switch off the the backstep when downhill, by using and {halfstepback}. The style {adaptglok} allows for a basic box relaxation to a stress-free state by choosing {relaxbox} {iso} or {axial}. One should give an approximated value for the bulk modulus {relaxbox_modulus} of the system. The default value of the relaxation rate {relaxbox_rate} should be optimum in most situations. to optimize the relaxation: {integrator}, {delaystep}, {dtgrow}, {dtshrink}, {alpha0}, {alphashrink}, {tmax} and {tmin}. Different Newton {integrator} can be selected: explicit Euler, semi-implicit Euler (= symplectic Euler), leapfrog, and velocity Verlet. The parameters {tmax} and {tmin} define the maximum and minimum timestep allowed during an adaptglok minimization. Those are not in time unit, but are multiplication factor applied to the "timestep"_timestep.html. For example, {tmax} = 2.0 means that the maximum value the timestep can reach is twice the value defined by "timestep"_timestep.html. Note that even with the default parameters being chosen to be reliable in most cases, adjusting "timestep"_timestep.html, {tmax} and {tmin} should be consider to optimize the minimization, in particular for large/complex system. For debugging purposes, it is possible to switch off the back step when downhill with {halfstepback} = 0. [Restrictions:] none Loading @@ -101,7 +108,7 @@ optimum in most situations. [Default:] The option defaults are dmax = 0.1, line = quadratic, integrator = eulerimplicit, delaystep = 20, dt_grow = 1.1, dt_shrink = 0.5, alpha0 = 0.25, alpha_shrink = 0.99, tmax = 2.0, tmin = 0.02, adaptstep = yes, halfstepback = yes, relaxbox = no, relaxbox_modulus = 1000000 and relaxbox_rate = 0.33. The option defaults are dmax = 0.1, line = quadratic, integrator = eulerimplicit, delaystep = 20, dt_grow = 1.1, dt_shrink = 0.5, alpha0 = 0.25, alpha_shrink = 0.99, tmax = 2.0, tmin = 0.02, adaptstep = yes and halfstepback = yes. doc/src/min_style.txt +2 −4 Changes for doc/src/min_style.txt: 2 added lines, 4 removed lines. Original line number Diff line number Diff line Loading @@ -66,10 +66,8 @@ fit the published and unpublished optimizations of the method described in "(Bitzek)"_#Bitzek. It includes a temporization of the variable timestep, a backstep when downhill and different integration schemes. Note that a basic scheme for having a stress-free simulation box is implemented within that style. Either the {quickmin} and {fire} styles are useful in the context of Either the {quickmin}, {fire} and {adaptglok} styles are useful in the context of nudged elastic band (NEB) calculations via the "neb"_neb.html command. NOTE: The damped dynamic minimizers use whatever timestep you have Loading @@ -77,7 +75,7 @@ defined via the "timestep"_timestep.html command. Often they will converge more quickly if you use a timestep about 10x larger than you would normally use for dynamics simulations. NOTE: The {quickmin}, {fire}, and {hftn} styles do not yet support the NOTE: The {quickmin}, {fire}, {adaptglok}, and {hftn} styles do not yet support the use of the "fix box/relax"_fix_box_relax.html command or minimizations involving the electron radius in "eFF"_pair_eff.html models. Loading doc/src/minimize.txt +3 −3 Changes for doc/src/minimize.txt: 3 added lines, 3 removed lines. Original line number Diff line number Diff line Loading @@ -64,9 +64,9 @@ backtracking method is described in Nocedal and Wright's Numerical Optimization (Procedure 3.1 on p 41). The "minimization styles"_min_style.html {quickmin}, {fire} and {adaptglok} perform damped dynamics using an Euler integration step (Velocity-Verlet for {adaptglok}). Thus they require a "timestep"_timestep.html be defined. {adaptglok} perform damped dynamics using an Euler integration step. The style {adaptglok} can also use a leapfrog or velocity Verlet integration step. Thus they require a "timestep"_timestep.html be defined. NOTE: The damped dynamic minimizers use whatever timestep you have defined via the "timestep"_timestep.html command. Often they will Loading src/min.cpp +9 −27 Changes for src/min.cpp: 9 added lines, 27 removed lines. Original line number Diff line number Diff line Loading @@ -56,17 +56,14 @@ Min::Min(LAMMPS *lmp) : Pointers(lmp) linestyle = 1; delaystep = 20; dt_grow = 1.1; dt_shrink = 0.5; dtgrow = 1.1; dtshrink = 0.5; alpha0 = 0.25; alpha_shrink = 0.99; alphashrink = 0.99; tmax = 2.0; tmin = 0.02; integrator = 0; relaxbox_modulus = 1000000; relaxbox_rate = 0.33; halfstepback_flag = 1; relaxbox_flag = 0; elist_global = elist_atom = NULL; vlist_global = vlist_atom = NULL; Loading Loading @@ -660,21 +657,21 @@ void Min::modify_params(int narg, char **arg) if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); delaystep = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"dt_grow") == 0) { } else if (strcmp(arg[iarg],"dtgrow") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); dt_grow = force->numeric(FLERR,arg[iarg+1]); dtgrow = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"dt_shrink") == 0) { } else if (strcmp(arg[iarg],"dtshrink") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); dt_shrink = force->numeric(FLERR,arg[iarg+1]); dtshrink = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"alpha0") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); alpha0 = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"alpha_shrink") == 0) { } else if (strcmp(arg[iarg],"alphashrink") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); alpha_shrink = force->numeric(FLERR,arg[iarg+1]); alphashrink = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"tmax") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); Loading @@ -690,21 +687,6 @@ void Min::modify_params(int narg, char **arg) else if (strcmp(arg[iarg+1],"no") == 0) halfstepback_flag = 0; else error->all(FLERR,"Illegal min_modify command"); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); if (strcmp(arg[iarg+1],"no") == 0) relaxbox_flag = 0; else if (strcmp(arg[iarg+1],"iso") == 0) relaxbox_flag = 1; else if (strcmp(arg[iarg+1],"axial") == 0) relaxbox_flag = 2; else error->all(FLERR,"Illegal min_modify command"); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox_modulus") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); relaxbox_modulus = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox_rate") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); relaxbox_rate = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"integrator") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); if (strcmp(arg[iarg+1],"eulerimplicit") == 0) integrator = 0; Loading src/min.h +7 −9 Changes for src/min.h: 7 added lines, 9 removed lines. Original line number Diff line number Diff line Loading @@ -58,15 +58,13 @@ class Min : protected Pointers { double dmax; // max dist to move any atom in one step int linestyle; // 0 = backtrack, 1 = quadratic, 2 = forcezero int delaystep; // minium steps of dynamics (adaptglok) double dt_grow,dt_shrink; // timestep increase, decrease (adaptglok) double alpha0,alpha_shrink; // mixing velocities+forces coefficient (adaptglok) double tmax,tmin; // timestep max, min (adaptglok) int integrator; // choose the style of time integrator (adaptglok) double relaxbox_modulus; // Bulk modulus used for box relax (for adaptglok) double relaxbox_rate; // for box relaxation to 0 pressure (for adaptglok) int halfstepback_flag; // 1: half step backward when v.f <= 0.0 (adaptglok) int relaxbox_flag; // 1: box relaxation iso; 2: axial // only for min_adaptglok int delaystep; // minium steps of dynamics double dtgrow,dtshrink; // timestep increase, decrease double alpha0,alphashrink; // mixing velocities+forces coefficient double tmax,tmin; // timestep multiplicators max, min int integrator; // Newton integration: euler, leapfrog, verlet... int halfstepback_flag; // half step backward when v.f <= 0.0 int nelist_global,nelist_atom; // # of PE,virial computes to check int nvlist_global,nvlist_atom; Loading Loading
doc/src/min_modify.txt +35 −28 Changes for doc/src/min_modify.txt: 35 added lines, 28 removed lines. Original line number Diff line number Diff line Loading @@ -19,19 +19,24 @@ keyword = {dmax} or {line} or {delaystep} or {dt_grow} or {dt_shrink} max = maximum distance for line search to move (distance units) {line} value = {backtrack} or {quadratic} or {forcezero} backtrack,quadratic,forcezero = style of linesearch to use {delaystep} value = {int} for adaptglok {dt_grow} value = {float} for adaptglok {dt_shrink} value = {float} for adaptglok {alpha0} value = {float} for adaptglok {alpha_shrink} value = {float} for adaptglok {tmax} value = {float} for adaptglok {tmin} value = {float} for adaptglok {delaystep} value = steps steps = number of steps for dynamics temporization, adaptglok only {dtgrow} value = grow grow = factor increasing the timestep, adaptglok only {dtshrink} value = shrink shrink = factor decreasing the timestep, adaptglok only {alpha0} value = alpha alpha = coefficient mixing velocities and forces, adaptglok only {alphashrink} value = shrink shrink = factor decreasing alpha, adaptglok only {tmax} value = coef coef = factor defining the maximum timestep, adaptglok only {tmin} value = coef coef = factor defining the minimum timestep, adaptglok only {integrator} value = {eulerimplicit} or {eulerexplicit} or {verlet} or {leapfrog} for adaptglok {halfstepback} value = {yes} or {no} for adaptglok :pre {relaxbox} value = {no} or {iso} or {axial} for adaptglok {relaxbox_modulus} value = {float} for adaptglok {relaxbox_rate} value = {float} for adaptglok or {leapfrog} eulerimplicit,eulerexplicit,verlet,leapfrog = integration scheme, adaptglok {halfstepback} value = {yes} or {no}, adaptglok only :pre :ule [Examples:] Loading Loading @@ -80,18 +85,20 @@ that difference may be smaller than machine epsilon even if atoms could move in the gradient direction to reduce forces further. The style {adaptglok} has several parameters that can be tuned in order to optimize the relaxation: {integrator}, {delaystep}, {dt_grow}, {dt_shrink}, {alpha0}, {alpha_shrink}, {tmax} and {tmin}. The default values have been chosen to be reliable in most cases, but one could be willing to adapt them for particular cases. In particular for debugging purpose, it is possible to switch off the the backstep when downhill, by using and {halfstepback}. The style {adaptglok} allows for a basic box relaxation to a stress-free state by choosing {relaxbox} {iso} or {axial}. One should give an approximated value for the bulk modulus {relaxbox_modulus} of the system. The default value of the relaxation rate {relaxbox_rate} should be optimum in most situations. to optimize the relaxation: {integrator}, {delaystep}, {dtgrow}, {dtshrink}, {alpha0}, {alphashrink}, {tmax} and {tmin}. Different Newton {integrator} can be selected: explicit Euler, semi-implicit Euler (= symplectic Euler), leapfrog, and velocity Verlet. The parameters {tmax} and {tmin} define the maximum and minimum timestep allowed during an adaptglok minimization. Those are not in time unit, but are multiplication factor applied to the "timestep"_timestep.html. For example, {tmax} = 2.0 means that the maximum value the timestep can reach is twice the value defined by "timestep"_timestep.html. Note that even with the default parameters being chosen to be reliable in most cases, adjusting "timestep"_timestep.html, {tmax} and {tmin} should be consider to optimize the minimization, in particular for large/complex system. For debugging purposes, it is possible to switch off the back step when downhill with {halfstepback} = 0. [Restrictions:] none Loading @@ -101,7 +108,7 @@ optimum in most situations. [Default:] The option defaults are dmax = 0.1, line = quadratic, integrator = eulerimplicit, delaystep = 20, dt_grow = 1.1, dt_shrink = 0.5, alpha0 = 0.25, alpha_shrink = 0.99, tmax = 2.0, tmin = 0.02, adaptstep = yes, halfstepback = yes, relaxbox = no, relaxbox_modulus = 1000000 and relaxbox_rate = 0.33. The option defaults are dmax = 0.1, line = quadratic, integrator = eulerimplicit, delaystep = 20, dt_grow = 1.1, dt_shrink = 0.5, alpha0 = 0.25, alpha_shrink = 0.99, tmax = 2.0, tmin = 0.02, adaptstep = yes and halfstepback = yes.
doc/src/min_style.txt +2 −4 Changes for doc/src/min_style.txt: 2 added lines, 4 removed lines. Original line number Diff line number Diff line Loading @@ -66,10 +66,8 @@ fit the published and unpublished optimizations of the method described in "(Bitzek)"_#Bitzek. It includes a temporization of the variable timestep, a backstep when downhill and different integration schemes. Note that a basic scheme for having a stress-free simulation box is implemented within that style. Either the {quickmin} and {fire} styles are useful in the context of Either the {quickmin}, {fire} and {adaptglok} styles are useful in the context of nudged elastic band (NEB) calculations via the "neb"_neb.html command. NOTE: The damped dynamic minimizers use whatever timestep you have Loading @@ -77,7 +75,7 @@ defined via the "timestep"_timestep.html command. Often they will converge more quickly if you use a timestep about 10x larger than you would normally use for dynamics simulations. NOTE: The {quickmin}, {fire}, and {hftn} styles do not yet support the NOTE: The {quickmin}, {fire}, {adaptglok}, and {hftn} styles do not yet support the use of the "fix box/relax"_fix_box_relax.html command or minimizations involving the electron radius in "eFF"_pair_eff.html models. Loading
doc/src/minimize.txt +3 −3 Changes for doc/src/minimize.txt: 3 added lines, 3 removed lines. Original line number Diff line number Diff line Loading @@ -64,9 +64,9 @@ backtracking method is described in Nocedal and Wright's Numerical Optimization (Procedure 3.1 on p 41). The "minimization styles"_min_style.html {quickmin}, {fire} and {adaptglok} perform damped dynamics using an Euler integration step (Velocity-Verlet for {adaptglok}). Thus they require a "timestep"_timestep.html be defined. {adaptglok} perform damped dynamics using an Euler integration step. The style {adaptglok} can also use a leapfrog or velocity Verlet integration step. Thus they require a "timestep"_timestep.html be defined. NOTE: The damped dynamic minimizers use whatever timestep you have defined via the "timestep"_timestep.html command. Often they will Loading
src/min.cpp +9 −27 Changes for src/min.cpp: 9 added lines, 27 removed lines. Original line number Diff line number Diff line Loading @@ -56,17 +56,14 @@ Min::Min(LAMMPS *lmp) : Pointers(lmp) linestyle = 1; delaystep = 20; dt_grow = 1.1; dt_shrink = 0.5; dtgrow = 1.1; dtshrink = 0.5; alpha0 = 0.25; alpha_shrink = 0.99; alphashrink = 0.99; tmax = 2.0; tmin = 0.02; integrator = 0; relaxbox_modulus = 1000000; relaxbox_rate = 0.33; halfstepback_flag = 1; relaxbox_flag = 0; elist_global = elist_atom = NULL; vlist_global = vlist_atom = NULL; Loading Loading @@ -660,21 +657,21 @@ void Min::modify_params(int narg, char **arg) if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); delaystep = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"dt_grow") == 0) { } else if (strcmp(arg[iarg],"dtgrow") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); dt_grow = force->numeric(FLERR,arg[iarg+1]); dtgrow = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"dt_shrink") == 0) { } else if (strcmp(arg[iarg],"dtshrink") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); dt_shrink = force->numeric(FLERR,arg[iarg+1]); dtshrink = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"alpha0") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); alpha0 = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"alpha_shrink") == 0) { } else if (strcmp(arg[iarg],"alphashrink") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); alpha_shrink = force->numeric(FLERR,arg[iarg+1]); alphashrink = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"tmax") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); Loading @@ -690,21 +687,6 @@ void Min::modify_params(int narg, char **arg) else if (strcmp(arg[iarg+1],"no") == 0) halfstepback_flag = 0; else error->all(FLERR,"Illegal min_modify command"); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); if (strcmp(arg[iarg+1],"no") == 0) relaxbox_flag = 0; else if (strcmp(arg[iarg+1],"iso") == 0) relaxbox_flag = 1; else if (strcmp(arg[iarg+1],"axial") == 0) relaxbox_flag = 2; else error->all(FLERR,"Illegal min_modify command"); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox_modulus") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); relaxbox_modulus = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"relaxbox_rate") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); relaxbox_rate = force->numeric(FLERR,arg[iarg+1]); iarg += 2; } else if (strcmp(arg[iarg],"integrator") == 0) { if (iarg+2 > narg) error->all(FLERR,"Illegal min_modify command"); if (strcmp(arg[iarg+1],"eulerimplicit") == 0) integrator = 0; Loading
src/min.h +7 −9 Changes for src/min.h: 7 added lines, 9 removed lines. Original line number Diff line number Diff line Loading @@ -58,15 +58,13 @@ class Min : protected Pointers { double dmax; // max dist to move any atom in one step int linestyle; // 0 = backtrack, 1 = quadratic, 2 = forcezero int delaystep; // minium steps of dynamics (adaptglok) double dt_grow,dt_shrink; // timestep increase, decrease (adaptglok) double alpha0,alpha_shrink; // mixing velocities+forces coefficient (adaptglok) double tmax,tmin; // timestep max, min (adaptglok) int integrator; // choose the style of time integrator (adaptglok) double relaxbox_modulus; // Bulk modulus used for box relax (for adaptglok) double relaxbox_rate; // for box relaxation to 0 pressure (for adaptglok) int halfstepback_flag; // 1: half step backward when v.f <= 0.0 (adaptglok) int relaxbox_flag; // 1: box relaxation iso; 2: axial // only for min_adaptglok int delaystep; // minium steps of dynamics double dtgrow,dtshrink; // timestep increase, decrease double alpha0,alphashrink; // mixing velocities+forces coefficient double tmax,tmin; // timestep multiplicators max, min int integrator; // Newton integration: euler, leapfrog, verlet... int halfstepback_flag; // half step backward when v.f <= 0.0 int nelist_global,nelist_atom; // # of PE,virial computes to check int nvlist_global,nvlist_atom; Loading
