Molecular Orbitals
Table of Contents
- 1. Context
- 2. Computation
- 2.1. Parameters of the cusp-correction functions
- 2.2. Computation of MOs: values only
- 2.3. Computation of MOs: values, gradient, Laplacian
- 2.4. Computation of MOs: Hessian
- 2.5. Computation of cusp-corrected MOs: values only
- 2.6. Computation of cusp-corrected MOs: values, gradient, Laplacian
- 2.7. Rescaling of MO coefficients
- 2.8. Test
1. Context
The following arrays are stored in the context:
mo_num |
Number of MOs | |
coefficient |
[mo_num][ao_num] |
MO coefficients |
coefficient_t |
[ao_num][mo_num] |
Transposed of the Orbital coefficients |
r_cusp |
[nucl_num] |
Radius of the functions for Cusp adjustments |
Computed data:
cusp_param |
[nucl_num][4][mo_num] |
Parameters of the functions for Cusp adjustments |
mo_value |
[point_num][mo_num] |
Value of the MOs at point positions |
mo_vgl |
[point_num][5][mo_num] |
Value, gradients, Laplacian of the MOs at point positions |
1.1. Data structure
The uninitialized integer contains one bit set to one for each
initialization function which has not been called. It becomes equal
to zero after all initialization functions have been called. The
struct is then initialized and provided == true.
Some values are initialized by default, and are not concerned by
this mechanism.
1.2. Initialization functions
To set the basis set, all the following functions need to be called.
qmckl_exit_code qmckl_set_mo_basis_mo_num (qmckl_context context, const int64_t mo_num); qmckl_exit_code qmckl_set_mo_basis_coefficient (qmckl_context context, const double * coefficient, const int64_t size_max); qmckl_exit_code qmckl_set_mo_basis_r_cusp (qmckl_context context, const double * r_cusp, const int64_t size_max);
When the basis set is completely entered, other data structures are
computed to accelerate the calculations.
The function qmckl_set_mo_basis_coefficients can be called multiple times to override the current coefficients in the context,
but the function qmckl_set_mo_basis_mo_num can only be called once in a context.
1.3. Cusp adjsutment functions
To activate the cusp adjustment, the user must enter the radius of the fitting function for each atom.
This function requires the computation of the value and gradients of the \(s\) AOs at the distance equal to the radius, and the values of the non-\(s\) AOs at the center.
1.4. Access functions
When all the data for the AOs have been provided, the following
function returns true.
bool qmckl_mo_basis_provided (const qmckl_context context);
1.4.1. Fortran interfaces
1.5. Update
It may be desirable to remove certain molecular orbitals (MOs) that do not significantly contribute to the wave function. In particular, in a single determinant calculation, the virtual MOs can be removed as they do not participate in the ground state configuration.
To select a subset of MOs that will be kept, an array of integers of
size mo_num can be created. If the integer corresponding to an MO is
zero, that MO is dropped and will not be included in the
calculation. If the integer is non-zero, the MO will be kept.
qmckl_exit_code qmckl_mo_basis_select_mo (const qmckl_context context, const int32_t* keep, const int64_t size_max);
1.5.1. Fortran interface
2. Computation
2.1. Parameters of the cusp-correction functions
2.2. Computation of MOs: values only
2.2.1. Get
qmckl_exit_code qmckl_get_mo_basis_mo_value(qmckl_context context, double* const mo_value, const int64_t size_max);
Uses the given array to compute the values.
qmckl_exit_code qmckl_get_mo_basis_mo_value_inplace (qmckl_context context, double* const mo_value, const int64_t size_max);
2.2.2. Provide
qmckl_exit_code qmckl_provide_mo_basis_mo_value(qmckl_context context);
qmckl_exit_code qmckl_provide_mo_basis_mo_value(qmckl_context context) { qmckl_exit_code rc = QMCKL_SUCCESS; if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) { return qmckl_failwith( context, QMCKL_INVALID_CONTEXT, "qmckl_provide_mo_basis_mo_value", NULL); } qmckl_context_struct* const ctx = (qmckl_context_struct*) context; assert (ctx != NULL); if (!ctx->mo_basis.provided) { return qmckl_failwith( context, QMCKL_NOT_PROVIDED, "qmckl_provide_mo_basis_mo_value", NULL); } /* Compute if necessary */ if (ctx->point.date > ctx->mo_basis.mo_value_date) { qmckl_memory_info_struct mem_info = qmckl_memory_info_struct_zero; mem_info.size = ctx->mo_basis.mo_num * ctx->point.num * sizeof(double); if (ctx->mo_basis.mo_value != NULL) { qmckl_memory_info_struct mem_info_test = qmckl_memory_info_struct_zero; rc = qmckl_get_malloc_info(context, ctx->mo_basis.mo_value, &mem_info_test); /* if rc != QMCKL_SUCCESS, we are maybe in an _inplace function because the memory was not allocated with qmckl_malloc */ if ((rc == QMCKL_SUCCESS) && (mem_info_test.size != mem_info.size)) { rc = qmckl_free(context, ctx->mo_basis.mo_value); assert (rc == QMCKL_SUCCESS); ctx->mo_basis.mo_value = NULL; } } /* Allocate array */ if (ctx->mo_basis.mo_value == NULL) { double* mo_value = (double*) qmckl_malloc(context, mem_info); if (mo_value == NULL) { return qmckl_failwith( context, QMCKL_ALLOCATION_FAILED, "qmckl_mo_basis_mo_value", NULL); } ctx->mo_basis.mo_value = mo_value; }
if (rc != QMCKL_SUCCESS) { return rc; } ctx->mo_basis.mo_value_date = ctx->date; } return QMCKL_SUCCESS; }
2.2.3. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
coefficient_t |
double[mo_num][ao_num] |
in | Transpose of the AO to MO transformation matrix |
ao_value |
double[point_num][ao_num] |
in | Value of the AOs |
mo_value |
double[point_num][mo_num] |
out | Value of the MOs |
The matrix of AO values is very sparse, so we use a sparse-dense matrix multiplication instead of a dgemm, as exposed in https://dx.doi.org/10.1007/978-3-642-38718-0_14.
function qmckl_compute_mo_basis_mo_value_doc(context, & ao_num, mo_num, point_num, & coefficient_t, ao_value, mo_value) & result(info) bind(C) use qmckl_constants implicit none integer (qmckl_context) , intent(in) , value :: context integer (c_int64_t) , intent(in) , value :: ao_num integer (c_int64_t) , intent(in) , value :: mo_num integer (c_int64_t) , intent(in) , value :: point_num real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(in) :: ao_value(ao_num,point_num) real (c_double ) , intent(out) :: mo_value(mo_num,point_num) integer(qmckl_exit_code) :: info integer*8 :: j,k info = QMCKL_SUCCESS do j=1,point_num mo_value(:,j) = 0.d0 do k=1,ao_num if (ao_value(k,j) == 0.d0) cycle mo_value(:,j) = mo_value(:,j) + coefficient_t(:,k) * ao_value(k,j) end do end do end function qmckl_compute_mo_basis_mo_value_doc
2.3. Computation of MOs: values, gradient, Laplacian
2.3.1. Get
qmckl_exit_code qmckl_get_mo_basis_mo_vgl(qmckl_context context, double* const mo_vgl, const int64_t size_max);
Uses the given array to compute the VGL.
qmckl_exit_code qmckl_get_mo_basis_mo_vgl_inplace (qmckl_context context, double* const mo_vgl, const int64_t size_max);
2.3.2. Provide
qmckl_exit_code qmckl_provide_mo_basis_mo_vgl(qmckl_context context);
qmckl_exit_code qmckl_provide_mo_basis_mo_vgl(qmckl_context context) { qmckl_exit_code rc = QMCKL_SUCCESS; if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) { return qmckl_failwith( context, QMCKL_INVALID_CONTEXT, "qmckl_provide_mo_basis_mo_vgl", NULL); } qmckl_context_struct* const ctx = (qmckl_context_struct*) context; assert (ctx != NULL); if (!ctx->mo_basis.provided) { return qmckl_failwith( context, QMCKL_NOT_PROVIDED, "qmckl_provide_mo_basis_mo_vgl", NULL); } /* Compute if necessary */ if (ctx->point.date > ctx->mo_basis.mo_vgl_date) { qmckl_memory_info_struct mem_info = qmckl_memory_info_struct_zero; mem_info.size = 5 * ctx->mo_basis.mo_num * ctx->point.num * sizeof(double); if (ctx->mo_basis.mo_vgl != NULL) { qmckl_memory_info_struct mem_info_test = qmckl_memory_info_struct_zero; rc = qmckl_get_malloc_info(context, ctx->mo_basis.mo_vgl, &mem_info_test); /* if rc != QMCKL_SUCCESS, we are maybe in an _inplace function because the memory was not allocated with qmckl_malloc */ if ((rc == QMCKL_SUCCESS) && (mem_info_test.size != mem_info.size)) { rc = qmckl_free(context, ctx->mo_basis.mo_vgl); assert (rc == QMCKL_SUCCESS); ctx->mo_basis.mo_vgl = NULL; } } /* Allocate array */ if (ctx->mo_basis.mo_vgl == NULL) { double* mo_vgl = (double*) qmckl_malloc(context, mem_info); if (mo_vgl == NULL) { return qmckl_failwith( context, QMCKL_ALLOCATION_FAILED, "qmckl_mo_basis_mo_vgl", NULL); } ctx->mo_basis.mo_vgl = mo_vgl; }
if (rc != QMCKL_SUCCESS) { return rc; } ctx->mo_basis.mo_vgl_date = ctx->date; } return QMCKL_SUCCESS; }
2.3.3. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
coefficient_t |
double[ao_num][mo_num] |
in | Transpose of the AO to MO transformation matrix |
ao_vgl |
double[point_num][5][ao_num] |
in | Value, gradients and Laplacian of the AOs |
mo_vgl |
double[point_num][5][mo_num] |
out | Value, gradients and Laplacian of the MOs |
The matrix of AO values is very sparse, so we use a sparse-dense matrix multiplication instead of a dgemm, as exposed in https://dx.doi.org/10.1007/978-3-642-38718-0_14.
function qmckl_compute_mo_basis_mo_vgl_doc(context, & ao_num, mo_num, point_num, & coefficient_t, ao_vgl, mo_vgl) & result(info) bind(C) use qmckl_constants implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: ao_num integer (c_int64_t) , intent(in) , value :: mo_num integer (c_int64_t) , intent(in) , value :: point_num real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(in) :: ao_vgl(ao_num,5,point_num) real (c_double ) , intent(out) :: mo_vgl(mo_num,5,point_num) integer(qmckl_exit_code) :: info integer*8 :: i,j,k double precision :: c1, c2, c3, c4, c5 info = QMCKL_SUCCESS do j=1,point_num mo_vgl(:,:,j) = 0.d0 do k=1,ao_num if (ao_vgl(k,1,j) /= 0.d0) then c1 = ao_vgl(k,1,j) c2 = ao_vgl(k,2,j) c3 = ao_vgl(k,3,j) c4 = ao_vgl(k,4,j) c5 = ao_vgl(k,5,j) do i=1,mo_num mo_vgl(i,1,j) = mo_vgl(i,1,j) + coefficient_t(i,k) * c1 mo_vgl(i,2,j) = mo_vgl(i,2,j) + coefficient_t(i,k) * c2 mo_vgl(i,3,j) = mo_vgl(i,3,j) + coefficient_t(i,k) * c3 mo_vgl(i,4,j) = mo_vgl(i,4,j) + coefficient_t(i,k) * c4 mo_vgl(i,5,j) = mo_vgl(i,5,j) + coefficient_t(i,k) * c5 end do end if end do end do end function qmckl_compute_mo_basis_mo_vgl_doc
2.4. Computation of MOs: Hessian
2.4.1. Get
qmckl_exit_code qmckl_get_mo_basis_mo_hessian(qmckl_context context, double* const mo_hessian, const int64_t size_max);
2.4.2. Provide
qmckl_exit_code qmckl_provide_mo_basis_mo_hessian(qmckl_context context);
qmckl_exit_code qmckl_provide_mo_basis_mo_hessian(qmckl_context context) { qmckl_exit_code rc = QMCKL_SUCCESS; if (qmckl_context_check(context) == QMCKL_NULL_CONTEXT) { return qmckl_failwith( context, QMCKL_INVALID_CONTEXT, "qmckl_provide_mo_basis_mo_hessian", NULL); } qmckl_context_struct* const ctx = (qmckl_context_struct*) context; assert (ctx != NULL); if (!ctx->mo_basis.provided) { return qmckl_failwith( context, QMCKL_NOT_PROVIDED, "qmckl_provide_mo_basis_mo_hessian", NULL); } /* Compute if necessary */ if (ctx->point.date > ctx->mo_basis.mo_hessian_date) { qmckl_memory_info_struct mem_info = qmckl_memory_info_struct_zero; mem_info.size = 3 * 3 * ctx->mo_basis.mo_num * ctx->point.num * sizeof(double); if (ctx->mo_basis.mo_hessian != NULL) { qmckl_memory_info_struct mem_info_test = qmckl_memory_info_struct_zero; rc = qmckl_get_malloc_info(context, ctx->mo_basis.mo_hessian, &mem_info_test); /* if rc != QMCKL_SUCCESS, we are maybe in an _inplace function because the memory was not allocated with qmckl_malloc */ if ((rc == QMCKL_SUCCESS) && (mem_info_test.size != mem_info.size)) { rc = qmckl_free(context, ctx->mo_basis.mo_hessian); assert (rc == QMCKL_SUCCESS); ctx->mo_basis.mo_hessian = NULL; } } /* Allocate array */ if (ctx->mo_basis.mo_hessian == NULL) { double* mo_hessian = (double*) qmckl_malloc(context, mem_info); if (mo_hessian == NULL) { return qmckl_failwith( context, QMCKL_ALLOCATION_FAILED, "qmckl_mo_basis_mo_hessian", NULL); } ctx->mo_basis.mo_hessian = mo_hessian; } rc = qmckl_provide_ao_basis_ao_hessian(context); if (rc != QMCKL_SUCCESS) { return qmckl_failwith( context, QMCKL_NOT_PROVIDED, "qmckl_ao_basis", NULL); } if (ctx->mo_basis.r_cusp == NULL) { /* No cusp correction */ rc = qmckl_compute_mo_basis_mo_hessian_doc_f(context, ctx->ao_basis.ao_num, ctx->mo_basis.mo_num, ctx->point.num, ctx->mo_basis.coefficient_t, ctx->ao_basis.ao_hessian, ctx->mo_basis.mo_hessian); } else { rc = qmckl_provide_en_distance(context); if (rc != QMCKL_SUCCESS) { return qmckl_failwith( context, QMCKL_NOT_PROVIDED, "qmckl_electron_en_distance", NULL); } rc = qmckl_compute_mo_basis_mo_hessian_cusp(context, ctx->nucleus.num, ctx->ao_basis.ao_num, ctx->mo_basis.mo_num, ctx->point.num, ctx->ao_basis.ao_nucl, ctx->ao_basis.ao_ang_mom, ctx->electron.en_distance, ctx->nucleus.coord, ctx->point.coord, ctx->mo_basis.r_cusp, ctx->mo_basis.cusp_param, ctx->mo_basis.coefficient_t, ctx->ao_basis.ao_hessian, ctx->mo_basis.mo_hessian); } if (rc != QMCKL_SUCCESS) { return rc; } ctx->mo_basis.mo_hessian_date = ctx->date; } return QMCKL_SUCCESS; }
2.4.3. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
coefficient_t |
double[mo_num][ao_num] |
in | Transpose of the AO to MO transformation matrix |
ao_hessian |
double[3][point_num][4][ao_num] |
in | Hessian of the AOs |
mo_hessian |
double[3][point_num][3][mo_num] |
out | Hessian of the MOs |
integer function qmckl_compute_mo_basis_mo_hessian_doc_f(context, & ao_num, mo_num, point_num, & coefficient_t, ao_hessian, mo_hessian) & bind(C) result(info) use, intrinsic :: iso_c_binding use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in), value :: ao_num, mo_num integer (c_int64_t) , intent(in), value :: point_num real (c_double ) , intent(in) :: ao_hessian(ao_num,4,point_num,3) real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(out) :: mo_hessian(mo_num,3,point_num,3) integer*8 :: i,j,k, kk, jj info = QMCKL_SUCCESS do jj=1,3 do j=1,point_num mo_hessian(:,:,j,jj) = 0.d0 do k=1,ao_num do kk=1,3 do i=1,mo_num mo_hessian(i,kk,j,jj) = mo_hessian(i,kk,j,jj) + coefficient_t(i,k) * ao_hessian(k,kk,j,jj) end do end do end do end do end do end function qmckl_compute_mo_basis_mo_hessian_doc_f
qmckl_exit_code qmckl_compute_mo_basis_mo_hessian_doc_f ( const qmckl_context context, const int64_t ao_num, const int64_t mo_num, const int64_t point_num, const double* coefficient_t, const double* ao_hessian, double* const mo_hessian );
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
nucl_num |
int64_t |
in | Number of nuclei |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
ao_nucl |
int64_t[ao_num] |
in | Nucleus on which the AO is centered |
ao_ang_mom |
int32_t[ao_num] |
in | Angular momentum of the shell |
en_distance |
double[point_num][nucl_num] |
in | Electron-nucleus distances |
nucl_coord |
double[3][nucl_num] |
in | Nuclear coordinates |
point_coord |
double[3][point_num] |
in | Electron coordinates |
r_cusp |
double[nucl_num] |
in | Cusp-adjustment radius |
cusp_param |
double[nucl_num][4][mo_num] |
in | Cusp-adjustment parameters |
coefficient_t |
double[mo_num][ao_num] |
in | Transpose of the AO to MO transformation matrix |
ao_hessian |
double[3][point_num][4][ao_num] |
in | Hessian of the AOs |
mo_hessian |
double[3][point_num][3][mo_num] |
out | Hessian of the MOs |
integer function qmckl_compute_mo_basis_mo_hessian_cusp_doc_f(context, & nucl_num, ao_num, mo_num, point_num, ao_nucl, ao_ang_mom, en_distance, & nucl_coord, point_coord, r_cusp, cusp_param, coefficient_t, ao_hessian, mo_hessian) & bind(C) result(info) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context) , intent(in), value :: context integer (c_int64_t) , intent(in), value :: nucl_num, ao_num, mo_num, point_num integer (c_int64_t) , intent(in) :: ao_nucl(ao_num) integer (c_int32_t) , intent(in) :: ao_ang_mom(ao_num) real (c_double ) , intent(in) :: en_distance(nucl_num, point_num) real (c_double ) , intent(in) :: nucl_coord(nucl_num,3) real (c_double ) , intent(in) :: point_coord(point_num,3) real (c_double ) , intent(in) :: r_cusp(nucl_num) real (c_double ) , intent(in) :: cusp_param(mo_num,4,nucl_num) real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(in) :: ao_hessian(ao_num,4,point_num,3) real (c_double ) , intent(out) :: mo_hessian(mo_num,3,point_num,3) integer*8 :: i,j,k, inucl, kk, jj double precision :: r, r_inv, r_vec(3) do jj=1,3 do j=1,point_num ! Initial contribution of the MO mo_hessian(:,:,j,jj) = 0.d0 do k=1,ao_num if (ao_hessian(k,1,j,jj) /= 0.d0) then inucl = ao_nucl(k)+1 if ( (en_distance(inucl,j) > r_cusp(inucl)) .or. (ao_ang_mom(k) > 0) ) then do kk=1,3 do i=1,mo_num mo_hessian(i,kk,j,jj) = mo_hessian(i,kk,j,jj) + coefficient_t(i,k) * ao_hessian(k,kk,j,jj) end do end do end if end if end do ! Cusp adjustment do inucl=1,nucl_num r = en_distance(inucl,j) if (r < r_cusp(inucl)) then r_vec(1:3) = point_coord(j,1:3) - nucl_coord(inucl,1:3) r_inv = 1.d0/r do kk=1,3 do i=1,mo_num mo_hessian(i,kk,j,jj) = mo_hessian(i,kk,j,jj) - & cusp_param(i,2,inucl) * r_vec(kk) * r_vec(jj) * r_inv**3 + & cusp_param(i,4,inucl) * 3 * r_vec(kk) * r_vec(jj) * r_inv if (kk == jj) then mo_hessian(i,kk,j,jj) = mo_hessian(i,kk,j,jj) + & cusp_param(i,2,inucl) * r_inv + & cusp_param(i,3,inucl) * 2 + & cusp_param(i,4,inucl) * 3 * r end if end do end do end if end do ! inucl end do end do ! jj info = QMCKL_SUCCESS end function qmckl_compute_mo_basis_mo_hessian_cusp_doc_f
qmckl_exit_code qmckl_compute_mo_basis_mo_hessian_cusp_doc_f ( const qmckl_context context, const int64_t nucl_num, const int64_t ao_num, const int64_t mo_num, const int64_t point_num, const int64_t* ao_nucl, const int32_t* ao_ang_mom, const double* en_distance, const double* nucl_coord, const double* point_coord, const double* r_cusp, const double* cusp_param, const double* coefficient_t, const double* ao_hessian, double* const mo_hessian );
qmckl_exit_code qmckl_compute_mo_basis_mo_hessian_cusp ( const qmckl_context context, const int64_t nucl_num, const int64_t ao_num, const int64_t mo_num, const int64_t point_num, const int64_t* ao_nucl, const int32_t* ao_ang_mom, const double* en_distance, const qmckl_matrix nucl_coord, const qmckl_matrix point_coord, const double* r_cusp, const qmckl_tensor cusp_param, const double* coefficient_t, const double* ao_hessian, double* const mo_hessian );
qmckl_exit_code qmckl_compute_mo_basis_mo_hessian_cusp (const qmckl_context context, const int64_t nucl_num, const int64_t ao_num, const int64_t mo_num, const int64_t point_num, const int64_t* ao_nucl, const int32_t* ao_ang_mom, const double* en_distance, const qmckl_matrix nucl_coord_matrix, const qmckl_matrix point_coord_matrix, const double* r_cusp, const qmckl_tensor cusp_param_tensor, const double* coefficient_t, const double* ao_hessian, double* const mo_hessian ) { qmckl_exit_code rc; double * nucl_coord = qmckl_alloc_double_of_matrix(context, nucl_coord_matrix); double * point_coord = qmckl_alloc_double_of_matrix(context, point_coord_matrix); double * cusp_param = qmckl_alloc_double_of_tensor(context, cusp_param_tensor); rc = qmckl_compute_mo_basis_mo_hessian_cusp_doc_f (context, nucl_num, ao_num, mo_num, point_num, ao_nucl, ao_ang_mom, en_distance, nucl_coord, point_coord, r_cusp, cusp_param, coefficient_t, ao_hessian, mo_hessian ); qmckl_free(context, nucl_coord); qmckl_free(context, point_coord); qmckl_free(context, cusp_param); return rc; }
2.5. Computation of cusp-corrected MOs: values only
2.5.1. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
nucl_num |
int64_t |
in | Number of nuclei |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
ao_nucl |
int64_t[ao_num] |
in | Nucleus on which the AO is centered |
ao_ang_mom |
int32_t[ao_num] |
in | Angular momentum of the shell |
en_distance |
double[point_num][nucl_num] |
in | Electron-nucleus distances |
r_cusp |
double[nucl_num] |
in | Cusp-adjustment radius |
cusp_param |
double[nucl_num][4][mo_num] |
in | Cusp-adjustment parameters |
coefficient_t |
double[ao_num][mo_num] |
in | Transpose of the AO to MO transformation matrix |
ao_value |
double[point_num][ao_num] |
in | Value of the AOs |
mo_value |
double[point_num][mo_num] |
out | Cusp correction for the values of the MOs |
function qmckl_compute_mo_basis_mo_value_cusp_doc(context, & nucl_num, ao_num, mo_num, point_num, ao_nucl, ao_ang_mom, en_distance, & r_cusp, cusp_param, coefficient_t, ao_value, mo_value) & result(info) bind(C) use qmckl_constants implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: ao_num integer (c_int64_t) , intent(in) , value :: mo_num integer (c_int64_t) , intent(in) , value :: point_num integer (c_int64_t) , intent(in) :: ao_nucl(ao_num) integer (c_int32_t) , intent(in) :: ao_ang_mom(ao_num) real (c_double ) , intent(in) :: en_distance(nucl_num,point_num) real (c_double ) , intent(in) :: r_cusp(nucl_num) real (c_double ) , intent(in) :: cusp_param(mo_num,4,nucl_num) real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(in) :: ao_value(ao_num,point_num) real (c_double ) , intent(out) :: mo_value(mo_num,point_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, k, inucl double precision :: r info = QMCKL_SUCCESS do i=1,point_num mo_value(:,i) = 0.d0 do k=1,ao_num if (ao_value(k,i) == 0.d0) cycle inucl = ao_nucl(k)+1 if ( (en_distance(inucl,i) < r_cusp(inucl)) .and. (ao_ang_mom(k) == 0) ) cycle mo_value(:,i) = mo_value(:,i) + coefficient_t(:,k) * ao_value(k,i) end do ! k do inucl=1,nucl_num r = en_distance(inucl,i) if (r > r_cusp(inucl)) cycle do j=1,mo_num mo_value(j,i) = mo_value(j,i) + & cusp_param(j,1,inucl) + r*(cusp_param(j,2,inucl) + r*( & cusp_param(j,3,inucl) + r* cusp_param(j,4,inucl) )) enddo enddo ! inucl enddo ! i end function qmckl_compute_mo_basis_mo_value_cusp_doc
2.6. Computation of cusp-corrected MOs: values, gradient, Laplacian
2.6.1. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
nucl_num |
int64_t |
in | Number of nuclei |
ao_num |
int64_t |
in | Number of AOs |
mo_num |
int64_t |
in | Number of MOs |
point_num |
int64_t |
in | Number of points |
ao_nucl |
int64_t[ao_num] |
in | Nucleus on which the AO is centered |
ao_ang_mom |
int32_t[ao_num] |
in | Angular momentum of the shell |
en_distance |
double[point_num][nucl_num] |
in | Electron-nucleus distances |
nucl_coord |
double[3][nucl_num] |
in | Nuclear coordinates |
point_coord |
double[3][point_num] |
in | Electron coordinates |
r_cusp |
double[nucl_num] |
in | Cusp-adjustment radius |
cusp_param |
double[nucl_num][4][mo_num] |
in | Cusp-adjustment parameters |
coefficient_t |
double[ao_num][mo_num] |
in | Transpose of the AO to MO transformation matrix |
ao_vgl |
double[point_num][5][ao_num] |
in | Value, gradients and Laplacian of the AOs |
mo_vgl |
double[point_num][5][mo_num] |
out | Value, gradients and Laplacian of the MOs |
function qmckl_compute_mo_basis_mo_vgl_cusp_doc(context, & nucl_num, ao_num, mo_num, point_num, ao_nucl, ao_ang_mom, en_distance, & nucl_coord, point_coord, r_cusp, cusp_param, coefficient_t, ao_vgl, mo_vgl) & result(info) bind(C) use qmckl_constants implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: ao_num integer (c_int64_t) , intent(in) , value :: mo_num integer (c_int64_t) , intent(in) , value :: point_num integer (c_int64_t) , intent(in) :: ao_nucl(ao_num) integer (c_int32_t) , intent(in) :: ao_ang_mom(ao_num) real (c_double ) , intent(in) :: en_distance(nucl_num,point_num) real (c_double ) , intent(in) :: nucl_coord(nucl_num,3) real (c_double ) , intent(in) :: point_coord(point_num,3) real (c_double ) , intent(in) :: r_cusp(nucl_num) real (c_double ) , intent(in) :: cusp_param(mo_num,4,nucl_num) real (c_double ) , intent(in) :: coefficient_t(mo_num,ao_num) real (c_double ) , intent(in) :: ao_vgl(ao_num,5,point_num) real (c_double ) , intent(out) :: mo_vgl(mo_num,5,point_num) integer(qmckl_exit_code) :: info integer*8 :: i,j,k, inucl double precision :: c1, c2, c3, c4, c5 double precision :: r, r_inv, r_vec(3) do j=1,point_num ! Initial contribution of the MO mo_vgl(:,:,j) = 0.d0 do k=1,ao_num if (ao_vgl(k,1,j) /= 0.d0) then inucl = ao_nucl(k)+1 if ( (en_distance(inucl,j) > r_cusp(inucl)) .or. (ao_ang_mom(k) > 0) ) then c1 = ao_vgl(k,1,j) c2 = ao_vgl(k,2,j) c3 = ao_vgl(k,3,j) c4 = ao_vgl(k,4,j) c5 = ao_vgl(k,5,j) do i=1,mo_num mo_vgl(i,1,j) = mo_vgl(i,1,j) + coefficient_t(i,k) * c1 mo_vgl(i,2,j) = mo_vgl(i,2,j) + coefficient_t(i,k) * c2 mo_vgl(i,3,j) = mo_vgl(i,3,j) + coefficient_t(i,k) * c3 mo_vgl(i,4,j) = mo_vgl(i,4,j) + coefficient_t(i,k) * c4 mo_vgl(i,5,j) = mo_vgl(i,5,j) + coefficient_t(i,k) * c5 end do end if end if end do ! Cusp adjustment do inucl=1,nucl_num r = en_distance(inucl,j) if (r < r_cusp(inucl)) then r_vec(1:3) = point_coord(j,1:3) - nucl_coord(inucl,1:3) r_inv = 1.d0/r do i=1,mo_num mo_vgl(i,1,j) = mo_vgl(i,1,j) + & cusp_param(i,1,inucl) + r*( & cusp_param(i,2,inucl) + r*( & cusp_param(i,3,inucl) + r*( & cusp_param(i,4,inucl) ))) c1 = r_inv * cusp_param(i,2,inucl) + 2.d0*cusp_param(i,3,inucl) + & r * 3.d0 * cusp_param(i,4,inucl) mo_vgl(i,2,j) = mo_vgl(i,2,j) + r_vec(1) * c1 mo_vgl(i,3,j) = mo_vgl(i,3,j) + r_vec(2) * c1 mo_vgl(i,4,j) = mo_vgl(i,4,j) + r_vec(3) * c1 mo_vgl(i,5,j) = mo_vgl(i,5,j) + & 2.d0*cusp_param(i,2,inucl)*r_inv + & 6.d0*cusp_param(i,3,inucl) + & 12.d0*cusp_param(i,4,inucl)*r end do end if end do ! inucl end do info = QMCKL_SUCCESS end function qmckl_compute_mo_basis_mo_vgl_cusp_doc
2.7. Rescaling of MO coefficients
When evaluating Slater determinants, the value of the determinants may get out of the range of double precision. A simple fix is to rescale the MO coefficients to put back the determinants in the correct range.
qmckl_exit_code qmckl_mo_basis_rescale(qmckl_context context, const double scaling_factor);