colvarbias_abf Class Reference

ABF bias. More...

#include <colvarbias_abf.h>

Inheritance diagram for colvarbias_abf:

List of all members.

Public Member Functions
	colvarbias_abf (std::string const &conf, char const *key)
	ABF bias constructor; parses the config file.
	~colvarbias_abf ()
	Destructor.
cvm::real	update ()

---

Detailed Description

ABF bias.

Definition at line 21 of file colvarbias_abf.h.

---

Constructor & Destructor Documentation

colvarbias_abf::colvarbias_abf (  std::string const &  conf, char const *  key )

ABF bias constructor; parses the config file. Definition at line 11 of file colvarbias_abf.C. References colvarmodule::fatal_error(), colvar_grid< T >::has_parent_data, colvarmodule::log(), colvar_grid_gradient::samples, and colvarmodule::temperature(). : colvarbias (conf, key), gradients (NULL), samples (NULL) { if (cvm::temperature() == 0.0) cvm::log ("WARNING: ABF should not be run without a thermostat or at 0 Kelvin!\n"); // ************* parsing general ABF options *********************** get_keyval (conf, "applyBias", apply_bias, true); if (!apply_bias) cvm::log ("WARNING: ABF biases will *not* be applied!\n"); get_keyval (conf, "updateBias", update_bias, true); if (!update_bias) cvm::log ("WARNING: ABF biases will *not* be updated!\n"); get_keyval (conf, "hideJacobian", hide_Jacobian, false); if (hide_Jacobian) { cvm::log ("Jacobian (geometric) forces will be handled internally.\n"); } else { cvm::log ("Jacobian (geometric) forces will be included in reported free energy gradients.\n"); } get_keyval (conf, "fullSamples", full_samples, 200); if ( full_samples <= 1 ) full_samples = 1; min_samples = full_samples / 2; // full_samples - min_samples >= 1 is guaranteed get_keyval (conf, "inputPrefix", input_prefix, std::vector<std::string> ()); get_keyval (conf, "outputFreq", output_freq, cvm::restart_out_freq); get_keyval (conf, "historyFreq", history_freq, 0); b_history_files = (history_freq > 0); // ************* checking the associated colvars ******************* if (colvars.size() == 0) { cvm::fatal_error ("Error: no collective variables specified for the ABF bias.\n"); } for (size_t i = 0; i < colvars.size(); i++) { if (colvars[i]->type() != colvarvalue::type_scalar) { cvm::fatal_error ("Error: ABF bias can only use scalar-type variables.\n"); } colvars[i]->enable (colvar::task_gradients); if (update_bias) { // Request calculation of system force (which also checks for availability) colvars[i]->enable (colvar::task_system_force); if (!colvars[i]->tasks[colvar::task_extended_lagrangian]) { // request computation of Jacobian force colvars[i]->enable (colvar::task_Jacobian_force); // request Jacobian force as part as system force // except if the user explicitly requires the "silent" Jacobian // correction AND the colvar has a single component if (hide_Jacobian) { if (colvars[i]->n_components() > 1) { cvm::log ("WARNING: colvar \"" + colvars[i]->name + "\" has multiple components; reporting its Jacobian forces\n"); colvars[i]->enable (colvar::task_report_Jacobian_force); } } else { colvars[i]->enable (colvar::task_report_Jacobian_force); } } } // Here we could check for orthogonality of the Cartesian coordinates // and make it just a warning if some parameter is set? } bin.assign (colvars.size(), 0); force_bin.assign (colvars.size(), 0); force = new cvm::real [colvars.size()]; // Construct empty grids based on the colvars samples = new colvar_grid_count (colvars); gradients = new colvar_grid_gradient (colvars); gradients->samples = samples; samples->has_parent_data = true; // If custom grids are provided, read them if ( input_prefix.size() > 0 ) { read_gradients_samples (); } cvm::log ("Finished ABF setup.\n"); }

colvarbias_abf::~colvarbias_abf (   )

Destructor. Definition at line 104 of file colvarbias_abf.C. { if (samples) { delete samples; samples = NULL; } if (gradients) { delete gradients; gradients = NULL; } delete [] force; }

---

Member Function Documentation

cvm::real colvarbias_abf::update (   )  [virtual]

Update the FE gradient, compute and apply biasing force also output data to disk if needed Implements colvarbias. Definition at line 123 of file colvarbias_abf.C. References colvar_grid_gradient::acc_force(), colvar_grid_gradient::average(), colvar_grid< T >::current_bin_scalar(), colvarmodule::debug(), colvar_grid< T >::index_ok(), colvarmodule::log(), colvarmodule::real, colvar_grid< T >::reset(), colvarmodule::step_absolute(), colvarmodule::step_relative(), and colvar_grid< T >::value(). { if (cvm::debug()) cvm::log ("Updating ABF bias " + this->name); if (cvm::step_relative() == 0) { // At first timestep, do only: // initialization stuff (file operations relying on n_abf_biases // compute current value of colvars if ( cvm::n_abf_biases == 1 && cvm::n_meta_biases == 0 ) { // This is the only ABF bias output_prefix = cvm::output_prefix; } else { output_prefix = cvm::output_prefix + "." + this->name; } for (size_t i=0; i<colvars.size(); i++) { bin[i] = samples->current_bin_scalar(i); } } else { for (size_t i=0; i<colvars.size(); i++) { bin[i] = samples->current_bin_scalar(i); } if ( update_bias && samples->index_ok (force_bin) ) { // Only if requested and within bounds of the grid... for (size_t i=0; i<colvars.size(); i++) { // get forces (lagging by 1 timestep) from colvars force[i] = colvars[i]->system_force(); } gradients->acc_force (force_bin, force); } } // save bin for next timestep force_bin = bin; // Reset biasing forces from previous timestep for (size_t i=0; i<colvars.size(); i++) { colvar_forces[i].reset(); } // Compute and apply the new bias, if applicable if ( apply_bias && samples->index_ok (bin) ) { size_t count = samples->value (bin); cvm::real fact = 1.0; // Factor that ensures smooth introduction of the force if ( count < full_samples ) { fact = ( count < min_samples) ? 0.0 : (cvm::real (count - min_samples)) / (cvm::real (full_samples - min_samples)); } const cvm::real * grad = &(gradients->value (bin)); if ( fact != 0.0 ) { if ( (colvars.size() == 1) && colvars[0]->periodic_boundaries() ) { // Enforce a zero-mean bias on periodic, 1D coordinates colvar_forces[0].real_value += fact * (grad[0] / cvm::real (count) - gradients->average ()); } else { for (size_t i=0; i<colvars.size(); i++) { // subtracting the mean force (opposite of the FE gradient) means adding the gradient colvar_forces[i].real_value += fact * grad[i] / cvm::real (count); // without .real_value, the above would do (cheap) runtime type checking } } } } if (output_freq && (cvm::step_absolute() % output_freq) == 0) { if (cvm::debug()) cvm::log ("ABF bias trying to write gradients and samples to disk"); write_gradients_samples (output_prefix); } if (b_history_files && (cvm::step_absolute() % history_freq) == 0) { // append to existing file only if cvm::step_absolute() > 0 // otherwise, backup and replace write_gradients_samples (output_prefix + ".hist", (cvm::step_absolute() > 0)); } return 0.0; // TODO compute bias energy whenever possible (i.e. 1D with updateBias off) }

---

The documentation for this class was generated from the following files:

• colvarbias_abf.h
• colvarbias_abf.C

---