source/densfromf/recursive_expansion/src/purification_general.h File Reference

Recursive density matrix expansion (or density matrix purification). More...

#include <iostream>
#include <fstream>
#include <sstream>
#include "matrix_typedefs.h"
#include "realtype.h"
#include "matrix_utilities.h"
#include "integral_matrix_wrappers.h"
#include "output.h"
#include "matrix_proxy.h"
#include "puri_info.h"
#include "constants.h"
#include "utilities.h"
#include "units.h"
#include "files_dense.h"
#include "files_sparse.h"
#include "files_sparse_bin.h"
#include "get_eigenvectors.h"

Go to the source code of this file.

Classes

class  PurificationGeneral< MatrixType >
 PurificationGeneral is an abstract class which provides an interface for SP2, SP2ACC and possibly other recursive expansions. More...

Defines

#define NUM_ADDITIONAL_ITERATIONS   0
#define PURI_OUTPUT_NNZ

Typedefs

typedef ergo_real real

Variables

real eucl_acc
 Tolerance used for computation of spectral norm.
real mixed_acc
 Tolerance used for computation of mixed norm.
real TOL_OVERLAPPING_BOUNDS
 If the difference between inner bounds for homo and lumo is less then this tolerance, i.e.
real THRESHOLD_EIG_TOLERANCE
 Inner homo and lumo bounds may be too good, and it may happen that computed eigenvalue slightly outside of given intervals.
int EIG_EMPTY
int EIG_SQUARE_INT
int EIG_PROJECTION_INT
int EIG_POWER_INT
int EIG_LANCZOS_INT

Detailed Description

Recursive density matrix expansion (or density matrix purification).

Author:
Anastasia Kruchinina responsible

Define Documentation

#define NUM_ADDITIONAL_ITERATIONS   0
#define PURI_OUTPUT_NNZ

Typedef Documentation

typedef ergo_real real

Variable Documentation

int EIG_EMPTY

Tolerance used for computation of spectral norm.

Referenced by PurificationGeneral< MatrixType >::purification_process().

Tolerance used for computation of mixed norm.

NOTE: If truncation is 0 this may not be enough, set to machine epsilon.

Referenced by PurificationGeneral< MatrixType >::purification_process().

Inner homo and lumo bounds may be too good, and it may happen that computed eigenvalue slightly outside of given intervals.

Thus we allow some flexibility for eigenvalue. Set threshold 1e-5 since otherwise for small molecules does not work.

Referenced by PurificationGeneral< MatrixType >::check_eigenvectors_at_the_end(), and PurificationGeneral< MatrixType >::get_interval_with_lambda().

If the difference between inner bounds for homo and lumo is less then this tolerance, i.e.

bounds are still bad, eigenvectors will not be computed. (Inner bounds are used to estimate iterations for computation of homo and lumo eigenvectors.)

Referenced by PurificationGeneral< MatrixType >::prepare_to_purification().


Generated on 10 Jul 2018 for ergo by  doxygen 1.6.1