CHAMP Jastrow Factor
Table of Contents
- 1. Introduction
- 2.
- 3. Context
- 4. Computation
- 4.1. Electron-electron component
- 4.1.1. Asymptotic component
- 4.1.2. Electron-electron rescaled distances
- 4.1.3. Electron-electron rescaled distance gradients and Laplacian with respect to electron coordinates
- 4.1.4. Electron-electron component
- 4.1.5. Derivative
- 4.1.6. Parameter Derivative of the Asymptotic component
- 4.1.7. Parameter Derivative
- 4.1.8. Parameter Derivative of the gradient and Laplacian
- 4.2. Electron-nucleus component
- 4.2.1. Asymptotic component
- 4.2.2. Electron-nucleus rescaled distances
- 4.2.3. Electron-electron rescaled distance gradients and Laplacian with respect to electron coordinates
- 4.2.4. Electron-nucleus component
- 4.2.5. Derivative
- 4.2.6. Parameter Derivative of the Asymptotic component
- 4.2.7. Parameter Derivative
- 4.2.8. Parameter Derivative of the gradient and Laplacian
- 4.3. Electron-electron-nucleus component
- 4.3.1. Electron-electron rescaled distances in \(J_\text{eeN}\)
- 4.3.2. Electron-electron rescaled distances derivatives in \(J_\text{eeN}\)
- 4.3.3. Electron-nucleus rescaled distances in \(J_\text{eeN}\)
- 4.3.4. Electron-nucleus rescaled distances derivatives in \(J_\text{eeN}\)
- 4.3.5. Temporary arrays for electron-electron-nucleus Jastrow \(f_{een}\)
- 4.3.6. Electron-electron-nucleus Jastrow \(f_{een}\)
- 4.3.7. Electron-electron-nucleus Jastrow \(f_{een}\) derivative
- 4.3.8. Electron-electron-nucleus Jastrow Parameter derivatives
- 4.3.9. Electron-electron-nucleus Parameter Derivative of the gradient and Laplacian
- 4.4. Total Jastrow
- 4.1. Electron-electron component
1. Introduction
The Jastrow factor depends on the electronic (\(\mathbf{r}\)) and nuclear (\(\mathbf{R}\)) coordinates. Its defined as \(\exp(J(\mathbf{r},\mathbf{R}))\), where
\[ J(\mathbf{r},\mathbf{R}) = J_{\text{eN}}(\mathbf{r},\mathbf{R}) + J_{\text{ee}}(\mathbf{r}) + J_{\text{eeN}}(\mathbf{r},\mathbf{R}) \]
In the following, we use the notations \(r_{ij} = |\mathbf{r}_i - \mathbf{r}_j|\) and \(R_{i\alpha} = |\mathbf{r}_i - \mathbf{R}_\alpha|\).
\(J_{\text{eN}}\) contains electron-nucleus terms:
\[ J_{\text{eN}}(\mathbf{r},\mathbf{R}) = \sum_{\alpha=1}^{N_\text{nucl}} \sum_{i=1}^{N_\text{elec}} \frac{a_{1\,\alpha}\, f_\alpha(R_{i\alpha})}{1+a_{2\,\alpha}\, f_\alpha(R_{i\alpha})} + \sum_{p=2}^{N_\text{ord}^a} a_{p+1\,\alpha}\, [f_\alpha(R_{i\alpha})]^p - J_{\text{eN}}^{\infty \alpha} \]
\(J_{\text{ee}}\) contains electron-electron terms: \[ J_{\text{ee}}(\mathbf{r}) = \sum_{i=1}^{N_\text{elec}} \sum_{j=1}^{i-1} \frac{\frac{1}{2}(1+\delta^{\uparrow\downarrow}_{ij}) b_1\, f_{\text{ee}}(r_{ij})}{1+b_2\, f_{\text{ee}}(r_{ij})} + \sum_{p=2}^{N_\text{ord}^b} b_{p+1}\, [f_{\text{ee}}(r_{ij})]^p - J_{ee}^\infty \]
and \(J_{\text{eeN}}\) contains electron-electron-Nucleus terms:
\[ J_{\text{eeN}}(\mathbf{r},\mathbf{R}) = \sum_{\alpha=1}^{N_{\text{nucl}}} \sum_{i=1}^{N_{\text{elec}}} \sum_{j=1}^{i-1} \sum_{p=2}^{N_{\text{ord}}} \sum_{k=0}^{p-1} \sum_{l=0}^{p-k-2\delta_{k,0}} c_{lkp\alpha} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ \left[ g_\alpha({R}_{i\alpha}) \right]^l + \left[ g_\alpha({R}_{j\alpha}) \right]^l \right] \left[ g_\alpha({R}_{i\,\alpha}) \, g_\alpha({R}_{j\alpha}) \right]^{(p-k-l)/2} \]
\(c_{lkp\alpha}\) are non-zero only when \(p-k-l\) is even.
\(f\) and \(g\) are scaling function defined as
\[ f_\alpha(r) = \frac{1-e^{-\kappa_\alpha\, r}}{\kappa_\alpha} \text{ and } g_\alpha(r) = e^{-\kappa_\alpha\, r} = 1-\kappa_\alpha f_\alpha(r). \]
The terms \(J_{\text{ee}}^\infty\) and \(J_{\text{eN}}^\infty\) are shifts to ensure that \(J_{\text{ee}}\) and \(J_{\text{eN}}\) have an asymptotic value of zero.
The eN and eeN parameters are the same of all identical nuclei. Warning: The types of nuclei use zero-based indexing.
The derivatives are computed with respect to the electron \(i\) for \[ r_{ij} = |r_i - r_j| \]
1.1. Reformulation of the three-body part
To accelerate the computation of the three-body part, the Jastrow factor is re-expressed as follows, with \(m=(p-k)/2 -l/2\):
\begin{eqnarray*} J_{kpl} & = & \sum_{\alpha=1}^{N_{\text{nucl}}} \sum_{i=1}^{N_{\text{elec}}} \sum_{j=1}^{i-1} c_{lkp\alpha} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ \left[ g_\alpha({R}_{i\alpha}) \right]^l + \left[ g_\alpha({R}_{j\alpha}) \right]^l \right] \left[ g_\alpha({R}_{i\,\alpha}) \, g_\alpha({R}_{j\alpha}) \right]^{m} \\ & = & \frac{1}{2} \sum_{\alpha=1}^{N_{\text{nucl}}} \sum_{i=1}^{N_{\text{elec}}} \sum_{j=1}^{N_{\text{elec}}} c_{lkp\alpha} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ \left[ g_\alpha({R}_{i\alpha}) \right]^l + \left[ g_\alpha({R}_{j\alpha}) \right]^l \right] \left[ g_\alpha({R}_{i\,\alpha}) \, g_\alpha({R}_{j\alpha}) \right]^{m} \\ & = & \frac{1}{2} \sum_{\alpha=1}^{N_{\text{nucl}}} \sum_{i=1}^{N_{\text{elec}}} \sum_{j=1}^{N_{\text{elec}}} c_{lkp\alpha} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ \left[ g_\alpha({R}_{i\alpha}) \right]^{l+m} \left[ g_\alpha({R}_{j\alpha}) \right]^{m} + \left[ g_\alpha({R}_{i\alpha}) \right]^{l} \left[ g_\alpha({R}_{j\alpha}) \right]^{l+m} \right] \\ & = & \sum_{\alpha=1}^{N_{\text{nucl}}} c_{lkp\alpha} \sum_{i=1}^{N_{\text{elec}}} \sum_{j=1}^{N_{\text{elec}}} \left[ g_\alpha({R}_{i\alpha}) \right]^{l+m} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ g_\alpha({R}_{j\alpha}) \right]^{m} \\ & = & \sum_{\alpha=1}^{N_{\text{nucl}}} c_{lkp\alpha} \sum_{i=1}^{N_{\text{elec}}} \left[ g_\alpha({R}_{i\alpha}) \right]^{l+m} P_{i\alpha}^{km}, \text{ with } P_{i\alpha}^{km} = \sum_{j=1}^{N_{\text{elec}}} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ g_\alpha({R}_{j\alpha}) \right]^{m}. \\ J & = & \sum_{p=2}^{N_{\text{ord}}} \sum_{k=0}^{p-1} \sum_{l=0}^{p-k-2\delta_{k,0}} \sum_{\alpha=1}^{N_{\text{nucl}}} c_{lkp\alpha} \sum_{i=1}^{N_{\text{elec}}} \left[ g_\alpha({R}_{i\alpha}) \right]^{(p-k+l)/2} P_{i\alpha}^{k, (p-k-l)/2} \end{eqnarray*}The computation of \(P\) scales as \(\mathcal{O}(N_\text{elec}^2 N_\text{nucl}n^2)\), and the computation of \(J\) scales as \(\mathcal{O}(N_\text{elec}N_\text{nucl}n^2)\).
2.
3. Context
The following data stored in the context:
| Variable | Type | Description |
|---|---|---|
uninitialized |
int32_t |
Keeps bits set for uninitialized data |
rescale_factor_ee |
double |
The distance scaling factor |
rescale_factor_en |
double[type_nucl_num] |
The distance scaling factor |
aord_num |
int64_t |
The number of a coeffecients |
bord_num |
int64_t |
The number of b coeffecients |
cord_num |
int64_t |
The number of c coeffecients |
type_nucl_num |
int64_t |
Number of Nuclei types |
type_nucl_vector |
int64_t[nucl_num] |
IDs of types of Nuclei. These use 0-based indexing as in C. |
a_vector |
double[aord_num + 1][type_nucl_num] |
a polynomial coefficients |
b_vector |
double[bord_num + 1] |
b polynomial coefficients |
c_vector |
double[dim_c_vector][type_nucl_num] |
c polynomial coefficients |
spin_independent |
int32_t |
If 1, use same parameters for parallel and anti-parallel spins. Otherwise, 0. |
Computed data:
| Variable | Type | In/Out | |
|---|---|---|---|
dim_c_vector |
int64_t |
Number of unique C coefficients | |
dim_c_vector_date |
uint64_t |
Last modification of the number of unique C coefficients | |
asymp_jasa |
double[type_nucl_num] |
Asymptotic component | |
asymp_jasa_date |
uint64_t |
Last modification of the asymptotic component | |
asymp_jasa_pderiv |
double[type_nucl_num][aord_num+1] |
Derivatives w.r.t. the A parameters of the Asymptotic component | |
asymp_jasa_pderiv_date |
uint64_t |
Last modification of the asymptotic component derivative | |
asymp_jasb |
double[2] |
Asymptotic component | |
asymp_jasb_date |
uint64_t |
Last modification of derivative of the asymptotic jastrow | |
asymp_jasb_pderiv |
double[2][bord_num+1] |
Derivative of the asymptotic jastrow w.r.t the jastrow parameters (up- or down-spin) | |
asymp_jasb_pderiv_date |
double[2][bord_num+1] |
Last modification of psrameter derivative of the asymptotic jastrow | |
c_vector_full |
double[dim_c_vector][nucl_num] |
vector of non-zero coefficients | |
c_vector_full_date |
uint64_t |
Keep track of changes here | |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
Transform l,k,p, and m into consecutive indices | |
lkpm_combined_index_date |
uint64_t |
Transform l,k,p, and m into consecutive indices | |
tmp_c |
double[walk_num][cord_num][cord_num+1][nucl_num][elec_num] |
vector of non-zero coefficients | |
dtmp_c |
double[walk_num][elec_num][4][nucl_num][cord_num+1][cord_num] |
vector of non-zero coefficients | |
ee_distance_rescaled |
double[walk_num][num][num] |
Electron-electron rescaled distances | |
ee_distance_rescaled_date |
uint64_t |
Last modification date of the electron-electron distances | |
ee_distance_rescaled_gl |
double[walk_num][num][num][4] |
Electron-electron rescaled distances derivatives | |
ee_distance_rescaled_gl_date |
uint64_t |
Last modification date of the electron-electron distance derivatives | |
en_distance_rescaled |
double[walk_num][nucl_num][num] |
Electron-nucleus distances | |
en_distance_rescaled_date |
uint64_t |
Last modification date of the electron-electron distances | |
en_distance_rescaled_gl |
double[walk_num][nucl_num][num][4] |
Electron-electron rescaled distances derivatives | |
en_distance_rescaled_gl_date |
uint64_t |
Last modification date of the electron-electron distance derivatives | |
een_rescaled_n |
double[walk_num][cord_num+1][nucl_num][elec_num] |
The electron-electron rescaled distances raised to the powers defined by cord | |
een_rescaled_n_date |
uint64_t |
Keep track of the date of creation | |
een_rescaled_e_gl |
double[walk_num][cord_num+1][elec_num][4][elec_num] |
The electron-electron rescaled distances raised to the powers defined by cord derivatives wrt electrons | |
een_rescaled_e_gl_date |
uint64_t |
Keep track of the date of creation | |
een_rescaled_n_gl |
double[walk_num][cord_num+1][nucl_num][4][elec_num] |
The electron-electron rescaled distances raised to the powers defined by cord derivatives wrt electrons | |
een_rescaled_n_gl_date |
uint64_t |
Keep track of the date of creation | |
factor_ee |
double[walk_num] |
Jastrow factor: electron-electron part | |
factor_ee_date |
uint64_t |
Jastrow factor: electron-electron part | |
factor_ee_pderiv |
double[bord_num+1] |
Parameter derivatives Jastrow factor: electron-electron part | |
factor_ee_pderiv_date |
uint64_t |
Parameter derivatives Jastrow factor: electron-electron part | |
factor_en |
double[walk_num] |
Jastrow factor: electron-nucleus part | |
factor_en_date |
uint64_t |
Jastrow factor: electron-nucleus part | |
factor_en_pderiv |
double[type_nucl_num][aord_num+1] |
Parameter derivatives Jastrow factor: electron-nucleus part | |
factor_en_pderiv_date |
uint64_t |
Parameter derivativesJastrow factor: electron-nucleus part | |
factor_een |
double[walk_num] |
Jastrow factor: electron-electron-nucleus part | |
factor_een_date |
uint64_t |
Jastrow factor: electron-electron-nucleus part | |
factor_een_pderiv |
double[type_nucl_num][dim_c_vector] |
Parameter derivatives Jastrow factor: electron-electron-nucleus part | |
factor_een_pderiv_date |
uint64_t |
Parameter derivatives Jastrow factor: electron-electron-nucleus part | |
factor_ee_gl |
double[walk_num][4][elec_num] |
Parameter derivatives Jastrow factor: electron-electron-nucleus part | |
factor_ee_gl_date |
uint64_t |
Keep track of the date for the derivative | |
factor_ee_gl_pderiv |
double[bord_num+1][elec_num][4] |
Parameter derivative of the gradient of the jastrow factor | Derivative of the gradient and Laplacian with respect to the Jastrow parameters |
factor_ee_gl_pderiv_date |
uint64_t |
Parameter derivative of the gradient of the jastrow factor | Derivative of the gradient and Laplacian with respect to the Jastrow parameters |
factor_en_gl |
double[walk_num][4][elec_num] |
Derivative of the Jastrow factor: electron-electron-nucleus part | |
factor_en_gl_date |
uint64_t |
Keep track of the date for the en derivative | |
factor_en_gl_pderiv |
double[type_nucl_num][aord_num+1][elec_num][4] |
Derivative of the Jastrow factor: electron-electron-nucleus part | |
factor_en_gl_pderiv_date |
uint64_t |
Keep track of the date for the en derivative | |
factor_een_gl |
double[walk_num][4][elec_num] |
Derivative of the Jastrow factor: electron-electron-nucleus part | |
factor_een_gl_date |
uint64_t |
Keep track of the date for the een derivative | |
factor_een_grad |
double[walk_num][3][elec_num] |
Gradient of the Jastrow factor: electron-electron-nucleus part | |
factor_een_grad_date |
uint64_t |
Keep track of the date for the een derivative | |
factor_een_gl_pderiv |
double[type_nucl_num][dim_c_vector][elec_num][4] |
Parameter Derivative of the Jastrow factor: electron-electron-nucleus part | |
factor_een_gl_pderiv_date |
uint64_t |
Keep track of the date for the en derivative | |
value |
double[walk_num] |
Value of the Jastrow factor | |
value_date |
uint64_t |
Keep track of the date | |
gl |
double[walk_num][4][elec_num] |
Gradient and Laplacian of the Jastrow factor | |
grad |
double[walk_num][3][elec_num] |
Gradient of the Jastrow factor | |
value_date |
uint64_t |
Keep track of the date |
3.1. 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.
3.2. Initialization functions
To prepare for the Jastrow and its derivative, all the following functions need to be called.
qmckl_exit_code qmckl_set_jastrow_champ_rescale_factor_ee (qmckl_context context, const double kappa_ee); qmckl_exit_code qmckl_set_jastrow_champ_rescale_factor_en (qmckl_context context, const double* kappa_en, const int64_t size_max); qmckl_exit_code qmckl_set_jastrow_champ_aord_num (qmckl_context context, const int64_t aord_num); qmckl_exit_code qmckl_set_jastrow_champ_bord_num (qmckl_context context, const int64_t bord_num); qmckl_exit_code qmckl_set_jastrow_champ_cord_num (qmckl_context context, const int64_t cord_num); qmckl_exit_code qmckl_set_jastrow_champ_type_nucl_num (qmckl_context context, const int64_t type_nucl_num); qmckl_exit_code qmckl_set_jastrow_champ_type_nucl_vector (qmckl_context context, const int64_t* type_nucl_vector, const int64_t size_max); qmckl_exit_code qmckl_set_jastrow_champ_a_vector (qmckl_context context, const double * a_vector, const int64_t size_max); qmckl_exit_code qmckl_set_jastrow_champ_b_vector (qmckl_context context, const double * b_vector, const int64_t size_max); qmckl_exit_code qmckl_set_jastrow_champ_c_vector (qmckl_context context, const double * c_vector, const int64_t size_max); qmckl_exit_code qmckl_set_jastrow_champ_spin_independent (qmckl_context context, const int32_t spin_independent);
When the required information is completely entered, other data structures are computed to accelerate the calculations. The intermediates factors are precontracted using BLAS LEVEL 3 operations.
3.2.0.1. Fortran interface
interface integer(qmckl_exit_code) function qmckl_set_jastrow_champ_rescale_factor_ee (context, & kappa_ee) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context real(c_double), intent(in), value :: kappa_ee end function qmckl_set_jastrow_champ_rescale_factor_ee integer(qmckl_exit_code) function qmckl_set_jastrow_champ_rescale_factor_en (context, & kappa_en, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(in) :: kappa_en(size_max) end function qmckl_set_jastrow_champ_rescale_factor_en integer(qmckl_exit_code) function qmckl_set_jastrow_champ_aord_num (context, & aord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: aord_num end function qmckl_set_jastrow_champ_aord_num integer(qmckl_exit_code) function qmckl_set_jastrow_champ_bord_num (context, & bord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: bord_num end function qmckl_set_jastrow_champ_bord_num integer(qmckl_exit_code) function qmckl_set_jastrow_champ_cord_num (context, & cord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: cord_num end function qmckl_set_jastrow_champ_cord_num integer(qmckl_exit_code) function qmckl_set_jastrow_champ_type_nucl_num (context, & type_nucl_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: type_nucl_num end function qmckl_set_jastrow_champ_type_nucl_num integer(qmckl_exit_code) function qmckl_set_jastrow_champ_type_nucl_vector (context, & type_nucl_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max integer(c_int64_t), intent(in) :: type_nucl_vector(size_max) end function qmckl_set_jastrow_champ_type_nucl_vector integer(qmckl_exit_code) function qmckl_set_jastrow_champ_a_vector(context, & a_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(in) :: a_vector(size_max) end function qmckl_set_jastrow_champ_a_vector integer(qmckl_exit_code) function qmckl_set_jastrow_champ_b_vector(context, & b_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(in) :: b_vector(size_max) end function qmckl_set_jastrow_champ_b_vector integer(qmckl_exit_code) function qmckl_set_jastrow_champ_c_vector(context, & c_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(in) :: c_vector(size_max) end function qmckl_set_jastrow_champ_c_vector integer(qmckl_exit_code) function qmckl_set_jastrow_champ_spin_independent(context, & spin_independent) bind(C) use, intrinsic :: iso_c_binding import implicit none integer(qmckl_context) , intent(in) , value :: context integer(c_int32_t), intent(in), value :: spin_independent end function qmckl_set_jastrow_champ_spin_independent end interface
3.3. Access functions
Along with these core functions, calculation of the jastrow factor requires the following additional information to be set:
When all the data for the AOs have been provided, the following
function returns true.
bool qmckl_jastrow_champ_provided (const qmckl_context context);
3.3.0.1. Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_rescale_factor_ee (context, & kappa_ee) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context real(c_double), intent(out) :: kappa_ee end function qmckl_get_jastrow_champ_rescale_factor_ee integer(qmckl_exit_code) function qmckl_get_jastrow_champ_rescale_factor_en (context, & kappa_en, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: kappa_en(size_max) end function qmckl_get_jastrow_champ_rescale_factor_en integer(qmckl_exit_code) function qmckl_get_jastrow_champ_aord_num (context, & aord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(out) :: aord_num end function qmckl_get_jastrow_champ_aord_num integer(qmckl_exit_code) function qmckl_get_jastrow_champ_bord_num (context, & bord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(out) :: bord_num end function qmckl_get_jastrow_champ_bord_num integer(qmckl_exit_code) function qmckl_get_jastrow_champ_cord_num (context, & cord_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(out) :: cord_num end function qmckl_get_jastrow_champ_cord_num integer(qmckl_exit_code) function qmckl_get_jastrow_champ_type_nucl_num (context, & type_nucl_num) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(out) :: type_nucl_num end function qmckl_get_jastrow_champ_type_nucl_num integer(qmckl_exit_code) function qmckl_get_jastrow_champ_type_nucl_vector (context, & type_nucl_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context), intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max integer(c_int64_t), intent(out) :: type_nucl_vector(size_max) end function qmckl_get_jastrow_champ_type_nucl_vector integer(qmckl_exit_code) function qmckl_get_jastrow_champ_a_vector(context, & a_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: a_vector(size_max) end function qmckl_get_jastrow_champ_a_vector integer(qmckl_exit_code) function qmckl_get_jastrow_champ_b_vector(context, & b_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: b_vector(size_max) end function qmckl_get_jastrow_champ_b_vector integer(qmckl_exit_code) function qmckl_get_jastrow_champ_c_vector(context, & c_vector, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in) , value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: c_vector(size_max) end function qmckl_get_jastrow_champ_c_vector integer(qmckl_exit_code) function qmckl_get_jastrow_champ_spin_independent(context, & spin_independent) bind(C) use, intrinsic :: iso_c_binding import implicit none integer(qmckl_context) , intent(in) , value :: context integer(c_int32_t), intent(out) :: spin_independent end function qmckl_get_jastrow_champ_spin_independent end interface
4. Computation
The computed data is stored in the context so that it can be reused by different kernels. To ensure that the data is valid, for each computed data the date of the context is stored when it is computed. To know if some data needs to be recomputed, we check if the date of the dependencies are more recent than the date of the data to compute. If it is the case, then the data is recomputed and the current date is stored.
4.1. Electron-electron component
4.1.1. Asymptotic component
Calculate the asymptotic component asymp_jasb to be subtracted from the
electron-electron jastrow factor \(J_{\text{ee}}\). Two values are
computed. The first one is for parallel spin pairs, and the
second one for antiparallel spin pairs.
If the spin_independent variable is set to 1, then
\(\delta^{\uparrow \downarrow}\) is always equal to one.
\[ J_{\text{ee}}^{\infty} = \frac{\frac{1}{2}(1+\delta^{\uparrow \downarrow})\,b_1 \kappa_\text{ee}^{-1}}{1 + b_2\, \kappa_\text{ee}^{-1}} + \sum_{p=2}^{N_\text{ord}^b} b_{p+1}\, \kappa_\text{ee}^{-p} \]
4.1.1.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_asymp_jasb(qmckl_context context, double* const asymp_jasb, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_asymp_jasb(context, & asymp_jasb, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: asymp_jasb(size_max) end function qmckl_get_jastrow_champ_asymp_jasb end interface
4.1.1.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
bord_num |
int64_t |
in | Order of the polynomial |
b_vector |
double[bord_num+1] |
in | Values of b |
rescale_factor_ee |
double |
in | Electron coordinates |
spin_independent |
int32_t |
in | If 1, same parameters for parallel and anti-parallel pairs |
asymp_jasb |
double[2] |
out | Asymptotic value |
function qmckl_compute_jastrow_champ_asymp_jasb_doc(context, & bord_num, b_vector, rescale_factor_ee, spin_independent, asymp_jasb) & bind(C) result(info) use qmckl implicit none integer (qmckl_context) , intent(in) , value :: context integer (c_int64_t) , intent(in) , value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) , value :: rescale_factor_ee integer (c_int32_t) , intent(in) , value :: spin_independent real (c_double ) , intent(out) :: asymp_jasb(2) integer(qmckl_exit_code) :: info integer*8 :: i, p double precision :: kappa_inv, x, asym_one kappa_inv = 1.0d0 / rescale_factor_ee info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_2 return endif asym_one = b_vector(1) * kappa_inv / (1.0d0 + b_vector(2) * kappa_inv) if (spin_independent == 1) then asymp_jasb(:) = (/asym_one, asym_one/) else asymp_jasb(:) = (/0.5d0*asym_one, asym_one/) end if x = kappa_inv do p = 2, bord_num x = x * kappa_inv do i = 1, 2 asymp_jasb(i) = asymp_jasb(i) + b_vector(p + 1) * x end do end do end function qmckl_compute_jastrow_champ_asymp_jasb_doc
4.1.2. Electron-electron rescaled distances
ee_distance_rescaled stores the matrix of the rescaled distances between all
pairs of electrons:
\[ C_{ij} = \frac{ 1 - e^{-\kappa r_{ij}}}{\kappa} \]
where \(r_{ij}\) is the matrix of electron-electron distances.
4.1.2.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_ee_distance_rescaled(qmckl_context context, double* const distance_rescaled, int64_t const size_max);
4.1.2.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
elec_num |
int64_t |
in | Number of electrons |
rescale_factor_ee |
double |
in | Factor to rescale ee distances |
walk_num |
int64_t |
in | Number of walkers |
coord |
double[3][walk_num][elec_num] |
in | Electron coordinates |
ee_distance |
double[walk_num][elec_num][elec_num] |
out | Electron-electron rescaled distances |
function qmckl_compute_ee_distance_rescaled_doc(context, & elec_num, rescale_factor_ee, walk_num, & coord, ee_distance_rescaled) & bind(C) result(info) use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: elec_num real (c_double ) , intent(in) , value :: rescale_factor_ee integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: coord(elec_num,walk_num,3) real (c_double ) , intent(out) :: ee_distance_rescaled(elec_num,elec_num,walk_num) integer(qmckl_exit_code) :: info integer*8 :: k info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif do k=1,walk_num info = qmckl_distance_rescaled(context, 'T', 'T', elec_num, elec_num, & coord(1,k,1), elec_num * walk_num, & coord(1,k,1), elec_num * walk_num, & ee_distance_rescaled(1,1,k), elec_num, rescale_factor_ee) if (info /= QMCKL_SUCCESS) then exit endif end do end function qmckl_compute_ee_distance_rescaled_doc
4.1.3. Electron-electron rescaled distance gradients and Laplacian with respect to electron coordinates
The rescaled distances, represented by \(C_{ij} = (1 - e^{-\kappa_\text{e} r_{ij}})/\kappa_\text{e}\)
are differentiated with respect to the electron coordinates.
This information is stored in the tensor
ee_distance_rescaled_gl. The initial three sequential
elements of this three-dimensional tensor provide the \(x\), \(y\), and \(z\)
direction derivatives, while the fourth index corresponds to the Laplacian.
4.1.3.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_ee_distance_rescaled_gl(qmckl_context context, double* const distance_rescaled_gl, const int64_t size_max);
4.1.3.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
elec_num |
int64_t |
in | Number of electrons |
rescale_factor_ee |
double |
in | Factor to rescale ee distances |
walk_num |
int64_t |
in | Number of walkers |
coord |
double[3][walk_num][elec_num] |
in | Electron coordinates |
ee_distance_rescaled_gl |
double[walk_num][elec_num][elec_num][4] |
out | Electron-electron rescaled distance derivatives |
function qmckl_compute_ee_distance_rescaled_gl_doc(context, & elec_num, rescale_factor_ee, walk_num, coord, ee_distance_rescaled_gl) & bind(C) result(info) use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: elec_num real (c_double ) , intent(in) , value :: rescale_factor_ee integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: coord(elec_num,walk_num,3) real (c_double ) , intent(out) :: ee_distance_rescaled_gl(4,elec_num,elec_num,walk_num) integer(qmckl_exit_code) :: info integer*8 :: k info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif do k=1,walk_num info = qmckl_distance_rescaled_gl(context, 'T', 'T', elec_num, elec_num, & coord(1,k,1), elec_num*walk_num, & coord(1,k,1), elec_num*walk_num, & ee_distance_rescaled_gl(1,1,1,k), elec_num, rescale_factor_ee) if (info /= QMCKL_SUCCESS) then exit endif end do end function qmckl_compute_ee_distance_rescaled_gl_doc
4.1.4. Electron-electron component
Calculate the electron-electron jastrow component factor_ee using the asymp_jasb
component and the electron-electron rescaled distances ee_distance_rescaled.
If the spin_independent variable is set to 1, then
\(\delta^{\uparrow \downarrow}\) is always equal to one.
\[ f_\text{ee} = \sum_{i,j
\(\delta\) is the spin factor, \(B\) is the vector of \(b\) parameters, \(C\) is the array of rescaled distances.
\(f_{\text{ee}}\) can be rewritten as:
\[ f_\text{ee} = \frac{1}{2} \left[ \sum_{i,j} \frac{\delta_{ij}^{\uparrow\downarrow} B_0\, C_{ij}}{1 + B_1\, C_{ij}} + \sum_{i,j} \sum_{k=2}^{n_\text{ord}} B_k\, C_{ij}^k \right] - \left[ \frac{n_\uparrow (n_\uparrow-1) + n_\downarrow (n_\downarrow-1)}{2}\, {J_{\text{ee}}^{\infty}}_{\uparrow \uparrow} + n_\uparrow\,n_\downarrow\, {J_{\text{ee}}^{\infty}}_{\uparrow \downarrow} \right] \]
4.1.4.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_ee(qmckl_context context, double* const factor_ee, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_ee (context, & factor_ee, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_ee(size_max) end function qmckl_get_jastrow_champ_factor_ee end interface
4.1.4.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
up_num |
int64_t |
in | Number of alpha electrons |
bord_num |
int64_t |
in | Number of coefficients |
b_vector |
double[bord_num+1] |
in | List of coefficients |
ee_distance_rescaled |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances |
asymp_jasb |
double[2] |
in | Asymptotic value of the Jastrow |
factor_ee |
double[walk_num] |
out | \(f_{ee}\) |
function qmckl_compute_jastrow_champ_factor_ee_doc(context, & walk_num, elec_num, up_num, bord_num, b_vector, & ee_distance_rescaled, asymp_jasb, spin_independent, factor_ee) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in), value :: walk_num integer (c_int64_t) , intent(in), value :: elec_num integer (c_int64_t) , intent(in), value :: up_num integer (c_int64_t) , intent(in), value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) :: ee_distance_rescaled(elec_num,elec_num,walk_num) real (c_double ) , intent(in) :: asymp_jasb(2) integer (c_int32_t) , intent(in), value :: spin_independent real (c_double ) , intent(out) :: factor_ee(walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, k, nw double precision :: x, xk info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif do nw =1, walk_num factor_ee(nw) = 0.0d0 do j=1,elec_num do i=1,j-1 x = ee_distance_rescaled(i,j,nw) if (spin_independent == 1) then factor_ee(nw) = factor_ee(nw) + b_vector(1) * x / (1.d0 + b_vector(2) * x) - asymp_jasb(2) else if ( (j <= up_num).or.(i > up_num) ) then factor_ee(nw) = factor_ee(nw) + 0.5d0 * b_vector(1) * x / (1.d0 + b_vector(2) * x) - asymp_jasb(1) else factor_ee(nw) = factor_ee(nw) + b_vector(1) * x / (1.d0 + b_vector(2) * x) - asymp_jasb(2) endif endif xk = x do k=2,bord_num xk = xk * x factor_ee(nw) = factor_ee(nw) + b_vector(k+1)* xk end do end do end do end do end function qmckl_compute_jastrow_champ_factor_ee_doc
4.1.5. Derivative
The derivative of factor_ee is computed using the ee_distance_rescaled and
the electron-electron rescaled distances derivatives
ee_distance_rescaled_gl.
There are four components, the gradient which has 3 components in the \(x, y, z\)
directions and the laplacian as the last component.
\[ \nabla_i f_\text{ee} = \sum_{j\ne i} \left[\frac{\delta_{ij}^{\uparrow\downarrow} B_0\, \nabla_i C_{ij}}{(1 + B_1\, C_{ij})^2} + \sum^{n_\text{ord}}_{k=2} B_k\, k\, C_{ij}^{k-1} \nabla C_{ij} \right] \]
\[ \Delta_i f_\text{ee} = \sum_{j \ne i} \left[ \delta_{ij}^{\uparrow\downarrow} B_0 \left(\frac{ \Delta_i C_{ij}}{(1 + B_1\, C_{ij})^2} -\frac{2\,B_1 \left(\nabla_i C_{ij}\right)^2 }{(1 + B_1\, C_{ij})^3} \right) + \sum^{n_\text{ord}}_{k=2} B_k\, k\, \left((k-1)\, C_{ij}^{k-2} \left(\nabla_i C_{ij}\right)^2 + C_{ij}^{k-1} \Delta_i C_{ij} \right) \right] \]
4.1.5.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_ee_gl(qmckl_context context, double* const factor_ee_gl, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_ee_gl (context, & factor_ee_gl, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_ee_gl(size_max) end function qmckl_get_jastrow_champ_factor_ee_gl end interface
4.1.5.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
up_num |
int64_t |
in | Number of alpha electrons |
bord_num |
int64_t |
in | Number of coefficients |
b_vector |
double[bord_num+1] |
in | List of coefficients |
ee_distance_rescaled |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances |
ee_distance_rescaled_gl |
double[walk_num][elec_num][elec_num][4] |
in | Electron-electron distances |
spin_independent |
int32_t |
in | If 1, same parameters for parallel and antiparallel spins |
factor_ee_gl |
double[walk_num][4][elec_num] |
out | Electron-electron distances |
function qmckl_compute_jastrow_champ_factor_ee_gl_doc( & context, walk_num, elec_num, up_num, bord_num, & b_vector, ee_distance_rescaled, ee_distance_rescaled_gl, & spin_independent, factor_ee_gl) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: up_num integer (c_int64_t) , intent(in) , value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) :: ee_distance_rescaled(elec_num,elec_num,walk_num) real (c_double ) , intent(in) :: ee_distance_rescaled_gl(4,elec_num,elec_num,walk_num) integer (c_int32_t) , intent(in) , value :: spin_independent real (c_double ) , intent(out) :: factor_ee_gl(elec_num,4,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, k, nw, ii double precision :: x, x1, kf double precision :: denom, invdenom, invdenom2, f double precision :: grad_c2 double precision :: dx(4) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif if ((spin_independent < 0).or.(spin_independent > 1)) then info = QMCKL_INVALID_ARG_8 return endif do nw =1, walk_num factor_ee_gl(:,:,nw) = 0.0d0 do j = 1, elec_num do i = 1, elec_num if (i == j) cycle x = ee_distance_rescaled(i,j,nw) denom = 1.0d0 + b_vector(2) * x invdenom = 1.0d0 / denom invdenom2 = invdenom * invdenom dx(1) = ee_distance_rescaled_gl(1, i, j, nw) dx(2) = ee_distance_rescaled_gl(2, i, j, nw) dx(3) = ee_distance_rescaled_gl(3, i, j, nw) dx(4) = ee_distance_rescaled_gl(4, i, j, nw) grad_c2 = dx(1)*dx(1) + dx(2)*dx(2) + dx(3)*dx(3) if (spin_independent == 1) then f = b_vector(1) * invdenom2 else if((i <= up_num .and. j <= up_num ) .or. (i > up_num .and. j > up_num)) then f = 0.5d0 * b_vector(1) * invdenom2 else f = b_vector(1) * invdenom2 end if end if factor_ee_gl(i,1,nw) = factor_ee_gl(i,1,nw) + f * dx(1) factor_ee_gl(i,2,nw) = factor_ee_gl(i,2,nw) + f * dx(2) factor_ee_gl(i,3,nw) = factor_ee_gl(i,3,nw) + f * dx(3) factor_ee_gl(i,4,nw) = factor_ee_gl(i,4,nw) & + f * (dx(4) - 2.d0 * b_vector(2) * grad_c2 * invdenom) kf = 2.d0 x1 = x x = 1.d0 do k=2, bord_num f = b_vector(k+1) * kf * x factor_ee_gl(i,1,nw) = factor_ee_gl(i,1,nw) + f * x1 * dx(1) factor_ee_gl(i,2,nw) = factor_ee_gl(i,2,nw) + f * x1 * dx(2) factor_ee_gl(i,3,nw) = factor_ee_gl(i,3,nw) + f * x1 * dx(3) factor_ee_gl(i,4,nw) = factor_ee_gl(i,4,nw) & + f * (x1 * dx(4) + (kf-1.d0) * grad_c2) x = x*x1 kf = kf + 1.d0 end do end do end do end do end function qmckl_compute_jastrow_champ_factor_ee_gl_doc
4.1.6. Parameter Derivative of the Asymptotic component
Calculate the derivatives of the asymptotic component w.r.t. the
jastrow parameters asymp_jasb_pderiv to be subtracted from the
electron-electron jastrow factor \(J_{\text{ee}}\) derivatives. Two values are
computed. The first one is for parallel spin pairs, and the
second one for antiparallel spin pairs.
If the spin_independent variable is set to 1, then
\(\delta^{\uparrow \downarrow}\) is always equal to one.
\[ J_{\text{ee}}^{\infty} = \frac{\frac{1}{2}(1+\delta^{\uparrow \downarrow})\,B_0 \kappa_\text{ee}^{-1}}{1 + B_1\, \kappa_\text{ee}^{-1}} + \sum_{p=2}^{N_\text{ord}^B} B_{p+1}\, \kappa_\text{ee}^{-p} \]
\[ \partial_{B_0} J_{\text{ee}}^\infty = \frac{\frac{1}{2}(1 + \delta^{\uparrow\downarrow}_{ij}) \kappa_{\text{ee}}^{-1}}{1 + B_1 \kappa_{\text{ee}}^{-1}} \]
\[ \partial_{B_1}J_{\text{ee}}^\infty = -\frac{ \frac{1}{2}(1 + \delta_{ij}) \kappa_{\text{ee}}^{-2}} {(1 + B_1 \kappa_{\text{ee}}^{-1})^2 } \]
\[ \partial_{B_k} J_{\text{ee}}^\infty = \kappa_{\text{ee}}^{-(k-1)} \] for \(k > 1\)
4.1.6.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_asymp_jasb_pderiv(qmckl_context context, double* const asymp_jasb_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_asymp_jasb_pderiv(context, & asymp_jasb_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: asymp_jasb_pderiv(size_max) end function qmckl_get_jastrow_champ_asymp_jasb_pderiv end interface
4.1.6.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
bord_num |
int64_t |
in | Order of the polynomial |
b_vector |
double[bord_num+1] |
in | Values of b |
rescale_factor_ee |
double |
in | Electron coordinates |
spin_independent |
int32_t |
in | If 1, same parameters for parallel and anti-parallel pairs |
asymp_jasb |
double[2][bord_num+1] |
out | Asymptotic value |
function qmckl_compute_jastrow_champ_asymp_jasb_pderiv_doc(context, & bord_num, b_vector, rescale_factor_ee, spin_independent, asymp_jasb_pderiv) & bind(C) result(info) use qmckl implicit none integer (qmckl_context) , intent(in) , value :: context integer (c_int64_t) , intent(in) , value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) , value :: rescale_factor_ee integer (c_int32_t) , intent(in) , value :: spin_independent real (c_double ) , intent(out) :: asymp_jasb_pderiv(bord_num+1,2) integer(qmckl_exit_code) :: info integer*8 :: i, p double precision :: kappa_inv, x, asym_one, asym_two kappa_inv = 1.0d0 / rescale_factor_ee info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_2 return endif asym_one = kappa_inv / (1.0d0 + b_vector(2) * kappa_inv) asym_two = -b_vector(1) * kappa_inv**2 / (1.0d0 + b_vector(2) * kappa_inv)**2 if (spin_independent == 1) then asymp_jasb_pderiv(1,:) = (/asym_one, asym_one/) asymp_jasb_pderiv(2,:) = (/asym_two, asym_two/) else asymp_jasb_pderiv(1,:) = (/0.5d0*asym_one, asym_one/) asymp_jasb_pderiv(2,:) = (/0.5d0*asym_two, asym_two/) end if x = kappa_inv do p = 2, bord_num x = x * kappa_inv do i = 1, 2 asymp_jasb_pderiv(p+1,i) = x end do end do end function qmckl_compute_jastrow_champ_asymp_jasb_pderiv_doc
4.1.7. Parameter Derivative
The derivative of factor_ee with respect to the jastrow parameters
b_vector is computed. These derivatives are computed fron the
ee_rescaled_distances and the parameter derivatives of the asymptotic part
asymp_jasb_pderiv. If spin_independent is set to 1, then
\(\delta^{\uparrow \downarrow}\) is always equal to 1.
\[ \partial_{B_0} f_\text{ee} = \sum_{j,i
\[ \partial_{B_1} f_\text{ee} = -\sum_{j,i
\[ \partial_{B_k} f_\text{ee} = \sum_{j, i
4.1.7.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_ee_pderiv(qmckl_context context, double* const factor_ee_pderiv, const int64_t size_max);
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_ee_pderiv (context, & factor_ee_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_ee_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_ee_pderiv end interface
4.1.7.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
up_num |
int64_t |
in | Number of alpha electrons |
bord_num |
int64_t |
in | Number of coefficients |
b_vector |
double[bord_num+1] |
in | List of coefficients |
ee_distance_rescaled |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances |
asymp_jasb_pderiv |
double[2][bord_num+1] |
in | Asymptotic value of the Jastrow |
spin_independent |
|||
factor_ee_pderiv |
double[bord_num+1] |
out | \(\partial_{B_i} f_{ee}\) |
function qmckl_compute_jastrow_champ_factor_ee_pderiv_doc(context, & walk_num, elec_num, up_num, bord_num, b_vector, & ee_distance_rescaled, asymp_jasb_pderiv, spin_independent, factor_ee_pderiv) & 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 :: walk_num integer (c_int64_t) , intent(in), value :: elec_num integer (c_int64_t) , intent(in), value :: up_num integer (c_int64_t) , intent(in), value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) :: ee_distance_rescaled(elec_num,elec_num,walk_num) real (c_double ) , intent(in) :: asymp_jasb_pderiv(bord_num+1, 2) integer (c_int32_t) , intent(in), value :: spin_independent real (c_double ) , intent(out) :: factor_ee_pderiv(bord_num+1) integer(qmckl_exit_code) :: info integer*8 :: i, j, k, nw double precision :: x, xk info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif factor_ee_pderiv = 0.0d0 do nw = 1, walk_num do j = 1, elec_num do i = 1, j - 1 x = ee_distance_rescaled(i,j,nw) if (spin_independent == 1) then factor_ee_pderiv(1) = factor_ee_pderiv(1) + x / (1.d0 + b_vector(2) * x) - asymp_jasb_pderiv(1,2) factor_ee_pderiv(2) = factor_ee_pderiv(2) - b_vector(1) * x**2 / (1.d0 + b_vector(2) * x)**2 - asymp_jasb_pderiv(2,2) else if ( (j <= up_num).or.(i > up_num) ) then factor_ee_pderiv(1) = factor_ee_pderiv(1) + 0.5d0 * x / (1.d0 + b_vector(2) * x) - asymp_jasb_pderiv(1,1) factor_ee_pderiv(2) = factor_ee_pderiv(2) - 0.5d0 * b_vector(1) * x**2 / (1.d0 + b_vector(2) * x)**2 & - asymp_jasb_pderiv(2,1) else factor_ee_pderiv(1) = factor_ee_pderiv(1) + x / (1.d0 + b_vector(2) * x) - asymp_jasb_pderiv(2,1) factor_ee_pderiv(2) = factor_ee_pderiv(2) - b_vector(1) * x**2 / (1.d0 + b_vector(2) * x)**2 & - asymp_jasb_pderiv(2,2) endif endif xk = x do k = 2, bord_num xk = xk * x factor_ee_pderiv(k+1) = factor_ee_pderiv(k+1) + xk - asymp_jasb_pderiv(k+1, 1) end do end do end do end do factor_ee_pderiv = factor_ee_pderiv / dble(walk_num) end function qmckl_compute_jastrow_champ_factor_ee_pderiv_doc
4.1.8. Parameter Derivative of the gradient and Laplacian
The derivative of factor_ee_gl with respect to all the jastrow parameters C is computed and stored into factor_ee_gl_pderiv
4.1.8.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_ee_gl_pderiv(qmckl_context context, double* const factor_ee_gl_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_ee_gl_pderiv (context, & factor_ee_gl_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_ee_gl_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_ee_gl_pderiv end interface
4.1.8.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
up_num |
int64_t |
in | Number of alpha electrons |
bord_num |
int64_t |
in | Number of coefficients |
b_vector |
double[bord_num+1] |
in | List of coefficients |
ee_distance_rescaled |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances |
ee_distance_rescaled_gl |
double[walk_num][elec_num][elec_num][4] |
in | Electron-electron distances |
spin_independent |
int32_t |
in | If 1, same parameters for parallel and antiparallel spins |
factor_ee_gl_pderiv |
double[4][elec_num][bord_num+1] |
out | Electron-electron distances |
function qmckl_compute_jastrow_champ_factor_ee_gl_pderiv_doc( & context, walk_num, elec_num, up_num, bord_num, & b_vector, ee_distance_rescaled, ee_distance_rescaled_gl, & spin_independent, factor_ee_gl_pderiv) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: up_num integer (c_int64_t) , intent(in) , value :: bord_num real (c_double ) , intent(in) :: b_vector(bord_num+1) real (c_double ) , intent(in) :: ee_distance_rescaled(elec_num,elec_num,walk_num) real (c_double ) , intent(in) :: ee_distance_rescaled_gl(4,elec_num,elec_num,walk_num) integer (c_int32_t) , intent(in) , value :: spin_independent real (c_double ) , intent(out) :: factor_ee_gl_pderiv(4,elec_num,bord_num+1) integer(qmckl_exit_code) :: info integer*8 :: i, j, k, nw, ii double precision :: x, x1, kf double precision :: denom, invdenom, invdenom2, invdenom3, f, f1, f2 double precision :: grad_c2 double precision :: dx(4) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (bord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif if ((spin_independent < 0).or.(spin_independent > 1)) then info = QMCKL_INVALID_ARG_8 return endif factor_ee_gl_pderiv(:,:,:) = 0.0d0 do nw = 1, walk_num do j = 1, elec_num do i = 1, elec_num if (i == j) cycle x = ee_distance_rescaled(i,j,nw) denom = 1.0d0 + b_vector(2) * x invdenom = 1.0d0 / denom invdenom2 = invdenom * invdenom invdenom3 = invdenom2 * invdenom dx(1) = ee_distance_rescaled_gl(1, i, j, nw) dx(2) = ee_distance_rescaled_gl(2, i, j, nw) dx(3) = ee_distance_rescaled_gl(3, i, j, nw) dx(4) = ee_distance_rescaled_gl(4, i, j, nw) grad_c2 = dx(1)*dx(1) + dx(2)*dx(2) + dx(3)*dx(3) if (spin_independent == 1) then f1 = invdenom2 f2 = -2.d0 * b_vector(1) * x * invdenom3 else if((i <= up_num .and. j <= up_num ) .or. (i > up_num .and. j > up_num)) then f1 = 0.5d0 * invdenom2 f2 = -1.d0 * b_vector(1) * x * invdenom3 else f1 = invdenom2 f2 = -2.d0 * b_vector(1) * x * invdenom3 end if end if factor_ee_gl_pderiv(1,i,1) = factor_ee_gl_pderiv(1,i,1) + f1 * dx(1) factor_ee_gl_pderiv(2,i,1) = factor_ee_gl_pderiv(2,i,1) + f1 * dx(2) factor_ee_gl_pderiv(3,i,1) = factor_ee_gl_pderiv(3,i,1) + f1 * dx(3) factor_ee_gl_pderiv(4,i,1) = factor_ee_gl_pderiv(4,i,1) + f1 * dx(4) & + 2.d0 * b_vector(2) * grad_c2 * f1 * invdenom factor_ee_gl_pderiv(1,i,2) = factor_ee_gl_pderiv(1,i,2) + f2 * dx(1) factor_ee_gl_pderiv(2,i,2) = factor_ee_gl_pderiv(2,i,2) + f2 * dx(2) factor_ee_gl_pderiv(3,i,2) = factor_ee_gl_pderiv(3,i,2) + f2 * dx(3) factor_ee_gl_pderiv(4,i,2) = factor_ee_gl_pderiv(4,i,2) & + 2.d0 * f1 * b_vector(1) * (-invdenom * x * dx(4) - invdenom2 * grad_c2 * (1 - 2.d0 * x * b_vector(2))) kf = 2.d0 x1 = x x = 1.d0 do k=3, bord_num + 1 f = kf * x factor_ee_gl_pderiv(1,i,k) = factor_ee_gl_pderiv(1,i,k) + f * x1 * dx(1) factor_ee_gl_pderiv(2,i,k) = factor_ee_gl_pderiv(2,i,k) + f * x1 * dx(2) factor_ee_gl_pderiv(3,i,k) = factor_ee_gl_pderiv(3,i,k) + f * x1 * dx(3) factor_ee_gl_pderiv(4,i,k) = factor_ee_gl_pderiv(4,i,k) & + f * (x1 * dx(4) + (kf-1.d0) * grad_c2) x = x*x1 kf = kf + 1.d0 end do end do end do end do factor_ee_gl_pderiv = factor_ee_gl_pderiv / dble(walk_num) end function qmckl_compute_jastrow_champ_factor_ee_gl_pderiv_doc
4.2. Electron-nucleus component
4.2.1. Asymptotic component
Calculate the asymptotic component asymp_jasa to be subtracted from the final
electron-nucleus jastrow factor \(J_{\text{eN}}\). The asymptotic component is calculated
via the a_vector and the electron-nucleus rescale factors rescale_factor_en.
\[ J_{\text{en}}^{\infty \alpha} = -\frac{a_1 \kappa_\alpha^{-1}}{1 + a_2 \kappa_\alpha^{-1}} \]
4.2.1.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_asymp_jasa(qmckl_context context, double* const asymp_jasa, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_asymp_jasa(context, & asymp_jasa, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: asymp_jasa(size_max) end function qmckl_get_jastrow_champ_asymp_jasa end interface
4.2.1.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
aord_num |
int64_t |
in | Order of the polynomial |
type_nucl_num |
int64_t |
in | Number of nucleus types |
a_vector |
double[type_nucl_num][aord_num+1] |
in | Values of a |
rescale_factor_en |
double[type_nucl_num] |
in | Electron nucleus distances |
asymp_jasa |
double[type_nucl_num] |
out | Asymptotic value |
function qmckl_compute_jastrow_champ_asymp_jasa(context, aord_num, type_nucl_num, a_vector, & rescale_factor_en, asymp_jasa) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: aord_num integer (c_int64_t) , intent(in) , value :: type_nucl_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: rescale_factor_en(type_nucl_num) real (c_double ) , intent(out) :: asymp_jasa(type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, p double precision :: kappa_inv, x, asym_one info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_2 return endif do i=1,type_nucl_num kappa_inv = 1.0d0 / rescale_factor_en(i) asymp_jasa(i) = a_vector(1,i) * kappa_inv / (1.0d0 + a_vector(2,i) * kappa_inv) x = kappa_inv do p = 2, aord_num x = x * kappa_inv asymp_jasa(i) = asymp_jasa(i) + a_vector(p+1, i) * x end do end do end function qmckl_compute_jastrow_champ_asymp_jasa
4.2.2. Electron-nucleus rescaled distances
en_distance_rescaled stores the matrix of the rescaled distances between
electrons and nuclei.
\[ C_{i\alpha} = \frac{ 1 - e^{-\kappa_\alpha R_{i\alpha}}}{\kappa_\alpha} \]
where \(R_{i\alpha}\) is the matrix of electron-nucleus distances.
4.2.2.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_en_distance_rescaled(qmckl_context context, double* const distance_rescaled, const int64_t size_max);
4.2.2.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of types of nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | Number of types of nuclei |
rescale_factor_en |
double[type_nucl_num] |
in | The factor for rescaled distances |
walk_num |
int64_t |
in | Number of walkers |
elec_coord |
double[3][walk_num][elec_num] |
in | Electron coordinates |
nucl_coord |
double[3][nucl_num] |
in | Nuclear coordinates |
en_distance_rescaled |
double[walk_num][nucl_num][elec_num] |
out | Electron-nucleus distances |
function qmckl_compute_en_distance_rescaled_doc(context, & elec_num, nucl_num, type_nucl_num, & type_nucl_vector, rescale_factor_en, walk_num, elec_coord, & nucl_coord, en_distance_rescaled) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) real (c_double ) , intent(in) :: rescale_factor_en(type_nucl_num) integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: elec_coord(elec_num,walk_num,3) real (c_double ) , intent(in) :: nucl_coord(nucl_num,3) real (c_double ) , intent(out) :: en_distance_rescaled(elec_num,nucl_num,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, k double precision :: coord(3) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_5 return endif do i=1, nucl_num coord(1:3) = nucl_coord(i,1:3) do k=1,walk_num info = qmckl_distance_rescaled(context, 'T', 'N', elec_num, 1_8, & elec_coord(1,k,1), elec_num*walk_num, coord, 3_8, & en_distance_rescaled(1,i,k), elec_num, rescale_factor_en(type_nucl_vector(i)+1)) if (info /= QMCKL_SUCCESS) then return endif end do end do end function qmckl_compute_en_distance_rescaled_doc
4.2.3. Electron-electron rescaled distance gradients and Laplacian with respect to electron coordinates
The rescaled distances, represented by \(C_{i\alpha} = (1 - e^{-\kappa_\alpha R_{i\alpha}})/\kappa\)
are differentiated with respect to the electron coordinates.
This information is stored in the tensor
en_distance_rescaled_gl. The initial three sequential
elements of this three-index tensor provide the \(x\), \(y\), and \(z\)
direction derivatives, while the fourth index corresponds to the Laplacian.
4.2.3.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_en_distance_rescaled_gl(qmckl_context context, double* const distance_rescaled_gl, const int64_t size_max);
4.2.3.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of nucleus types |
type_nucl_vector |
int64_t[nucl_num] |
in | Array of nucleus types |
rescale_factor_en |
double[nucl_num] |
in | The factors for rescaled distances |
walk_num |
int64_t |
in | Number of walkers |
elec_coord |
double[3][walk_num][elec_num] |
in | Electron coordinates |
nucl_coord |
double[3][nucl_num] |
in | Nuclear coordinates |
en_distance_rescaled_gl |
double[walk_num][nucl_num][elec_num][4] |
out | Electron-nucleus distance derivatives |
function qmckl_compute_en_distance_rescaled_gl_doc(context, elec_num, nucl_num, & type_nucl_num, type_nucl_vector, rescale_factor_en, walk_num, elec_coord, & nucl_coord, en_distance_rescaled_gl) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in) :: context integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) real (c_double ) , intent(in) :: rescale_factor_en(nucl_num) integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: elec_coord(elec_num,walk_num,3) real (c_double ) , intent(in) :: nucl_coord(nucl_num,3) real (c_double ) , intent(out) :: en_distance_rescaled_gl(4,elec_num,nucl_num,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, k double precision :: coord(3) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_5 return endif do i=1, nucl_num coord(1:3) = nucl_coord(i,1:3) do k=1,walk_num info = qmckl_distance_rescaled_gl(context, 'T', 'T', elec_num, 1_8, & elec_coord(1,k,1), elec_num*walk_num, coord, 1_8, & en_distance_rescaled_gl(1,1,i,k), elec_num, rescale_factor_en(type_nucl_vector(i)+1)) if (info /= QMCKL_SUCCESS) then return endif end do end do end function qmckl_compute_en_distance_rescaled_gl_doc
4.2.4. Electron-nucleus component
Calculate the electron-electron jastrow component factor_en using the a_vector
coeffecients and the electron-nucleus rescaled distances en_distance_rescaled.
\[ f_{\alpha}(R_{i\alpha}) = - \sum_{i,j
4.2.4.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_en(qmckl_context context, double* const factor_en, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_en (context, & factor_en, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_en(size_max) end function qmckl_get_jastrow_champ_factor_en end interface
4.2.4.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of unique nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | IDs of unique nuclei |
aord_num |
int64_t |
in | Number of coefficients |
a_vector |
double[type_nucl_num][aord_num+1] |
in | List of coefficients |
en_distance_rescaled |
double[walk_num][nucl_num][elec_num] |
in | Electron-nucleus distances |
asymp_jasa |
double[type_nucl_num] |
in | Type of nuclei |
factor_en |
double[walk_num] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_en_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, & type_nucl_vector, aord_num, a_vector, & en_distance_rescaled, asymp_jasa, factor_en) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: aord_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: en_distance_rescaled(elec_num,nucl_num,walk_num) real (c_double ) , intent(in) :: asymp_jasa(type_nucl_num) real (c_double ) , intent(out) :: factor_en(walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, p, nw double precision :: x, power_ser info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (type_nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_7 return endif do nw =1, walk_num factor_en(nw) = 0.0d0 do a = 1, nucl_num do i = 1, elec_num x = en_distance_rescaled(i, a, nw) factor_en(nw) = factor_en(nw) + a_vector(1, type_nucl_vector(a)+1) * x / & (1.0d0 + a_vector(2, type_nucl_vector(a)+1) * x) - asymp_jasa(type_nucl_vector(a)+1) do p = 2, aord_num x = x * en_distance_rescaled(i, a, nw) factor_en(nw) = factor_en(nw) + a_vector(p + 1, type_nucl_vector(a)+1) * x end do end do end do end do end function qmckl_compute_jastrow_champ_factor_en_doc
4.2.5. Derivative
Calculate the electron-electron jastrow component factor_en_gl derivative
with respect to the electron coordinates using the en_distance_rescaled and
en_distance_rescaled_gl which are already calculated previously.
The derivative is calculated in the function qmckl_compute_jastrow_champ_factor_en_gl.
The formula is given by:
\[
\nabla_i f_{\alpha}(R_{i\alpha}) = \sum_{j=1}^{N_\text{elec}} \left[
\frac{ A_0\, \nabla_i C_{ij} }{(1 + A_1 C_{ij})^2} +
\sum_{k=2}^{N^\alpha_{\text{ord}}} A_k\, k\, C_{ij}^{k-1}\,\nabla_i C_{ij} \right]
\]
\[ \Delta_i f_{\alpha}(R_{i\alpha}) = \sum_{j=1}^{N_\text{elec}} \left[ \frac{ A_0\, \Delta_i C_{ij} }{(1 + A_1 C_{ij})^2} - \frac{ 2 A_0\, A_1 (\nabla_i C_{ij})^2}{(1 + A_1 C_{ij})^3} + \sum_{k=2}^{N^\alpha_{\text{ord}}} A_k\, k\, C_{ij}^{k-1} C_{ij}^{k-2} \left[ \Delta_i C_{ij} + (k-1)(\nabla_i C_{ij})^2 \right] \right] \]
4.2.5.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_en_gl(qmckl_context context, double* const factor_en_gl, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_en_gl (context, & factor_en_gl, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_en_gl(size_max) end function qmckl_get_jastrow_champ_factor_en_gl end interface
4.2.5.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of unique nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | IDs of unique nuclei |
aord_num |
int64_t |
in | Number of coefficients |
a_vector |
double[type_nucl_num][aord_num+1] |
in | List of coefficients |
en_distance_rescaled |
double[walk_num][nucl_num][elec_num] |
in | Electron-nucleus distances |
en_distance_rescaled_gl |
double[walk_num][nucl_num][elec_num][4] |
in | Electron-nucleus distance derivatives |
factor_en_gl |
double[walk_num][4][elec_num] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_en_gl_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, & type_nucl_vector, aord_num, a_vector, & en_distance_rescaled, en_distance_rescaled_gl, factor_en_gl) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: aord_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: en_distance_rescaled(elec_num,nucl_num,walk_num) real (c_double ) , intent(in) :: en_distance_rescaled_gl(4, elec_num,nucl_num,walk_num) real (c_double ) , intent(out) :: factor_en_gl(elec_num,4,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, k, nw, ii double precision :: x, x1, kf double precision :: denom, invdenom, invdenom2, f double precision :: grad_c2 double precision :: dx(4) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_7 return endif do nw =1, walk_num factor_en_gl(:,:,nw) = 0.0d0 do a = 1, nucl_num do i = 1, elec_num x = en_distance_rescaled(i,a,nw) if(abs(x) < 1.d-12) continue denom = 1.0d0 + a_vector(2, type_nucl_vector(a)+1) * x invdenom = 1.0d0 / denom invdenom2 = invdenom*invdenom dx(1) = en_distance_rescaled_gl(1,i,a,nw) dx(2) = en_distance_rescaled_gl(2,i,a,nw) dx(3) = en_distance_rescaled_gl(3,i,a,nw) dx(4) = en_distance_rescaled_gl(4,i,a,nw) f = a_vector(1, type_nucl_vector(a)+1) * invdenom2 grad_c2 = dx(1)*dx(1) + dx(2)*dx(2) + dx(3)*dx(3) factor_en_gl(i,1,nw) = factor_en_gl(i,1,nw) + f * dx(1) factor_en_gl(i,2,nw) = factor_en_gl(i,2,nw) + f * dx(2) factor_en_gl(i,3,nw) = factor_en_gl(i,3,nw) + f * dx(3) factor_en_gl(i,4,nw) = factor_en_gl(i,4,nw) & + f * (dx(4) - 2.d0 * a_vector(2, type_nucl_vector(a)+1) * grad_c2 * invdenom) kf = 2.d0 x1 = x x = 1.d0 do k=2, aord_num f = a_vector(k+1,type_nucl_vector(a)+1) * kf * x factor_en_gl(i,1,nw) = factor_en_gl(i,1,nw) + f * x1 * dx(1) factor_en_gl(i,2,nw) = factor_en_gl(i,2,nw) + f * x1 * dx(2) factor_en_gl(i,3,nw) = factor_en_gl(i,3,nw) + f * x1 * dx(3) factor_en_gl(i,4,nw) = factor_en_gl(i,4,nw) & + f * (x1 * dx(4) + (kf-1.d0) * grad_c2) x = x*x1 kf = kf + 1.d0 end do end do end do end do end function qmckl_compute_jastrow_champ_factor_en_gl_doc
4.2.5.3. Test
4.2.6. Parameter Derivative of the Asymptotic component
Calculate the derivative of the asymptotic component asymp_jasa to be subtracted from the final
electron-nucleus jastrow factor parameter derivatives \(J_{\text{eN}}\). The asymptotic component is calculated
via the a_vector and the electron-nucleus rescale factors rescale_factor_en.
\[ J_{\text{en}}^{\infty} = \frac{A_0 \kappa_\text{en}^{-1}}{1 + A_1\, \kappa_\text{en}^{-1}} + \sum_{p=2}^{N_\text{ord}^B} A_{p+1}\, \kappa_\text{en}^{-p} \]
\[ \partial_{B_0} J_{\text{ee}}^\infty = \frac{\frac{1}{2}(1 + \delta^{\uparrow\downarrow}_{ij}) \kappa_{\text{ee}}^{-1}}{1 + B_1 \kappa_{\text{ee}}^{-1}} \]
\[ \partial_{B_1}J_{\text{ee}}^\infty = -\frac{ \frac{1}{2}(1 + \delta_{ij}) \kappa_{\text{ee}}^{-2}} {(1 + B_1 \kappa_{\text{ee}}^{-1})^2 } \]
\[ \partial_{B_k} J_{\text{ee}}^\infty = \kappa_{\text{ee}}^{-(k-1)} \] for \(k > 1\)
\[ J_{\text{en}}^{\infty \alpha} = -\frac{a_1 \kappa_\alpha^{-1}}{1 + a_2 \kappa_\alpha^{-1}} \]
4.2.6.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_asymp_jasa_pderiv(qmckl_context context, double* const asymp_jasa_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_asymp_jasa_pderiv(context, & asymp_jasa_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: asymp_jasa_pderiv(size_max) end function qmckl_get_jastrow_champ_asymp_jasa_pderiv end interface
4.2.6.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
aord_num |
int64_t |
in | Order of the polynomial |
type_nucl_num |
int64_t |
in | Number of nucleus types |
a_vector |
double[type_nucl_num][aord_num+1] |
in | Values of a |
rescale_factor_en |
double[type_nucl_num] |
in | Electron nucleus distances |
asymp_jasa_pderiv |
double[type_nucl_num][aord_num+1] |
out | Parameter derivative of Asymptotic value |
function qmckl_compute_jastrow_champ_asymp_jasa_pderiv(context, aord_num, type_nucl_num, a_vector, & rescale_factor_en, asymp_jasa_pderiv) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: aord_num integer (c_int64_t) , intent(in) , value :: type_nucl_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: rescale_factor_en(type_nucl_num) real (c_double ) , intent(out) :: asymp_jasa_pderiv(aord_num+1, type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, p double precision :: kappa_inv, x, asym_one info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_2 return endif do i=1,type_nucl_num kappa_inv = 1.0d0 / rescale_factor_en(i) asymp_jasa_pderiv(1,i) = kappa_inv / (1.0d0 + a_vector(2,i) * kappa_inv) asymp_jasa_pderiv(2,i) = - a_vector(1,i) * kappa_inv**2 / (1.0d0 + a_vector(2,i) * kappa_inv)**2 x = kappa_inv do p = 2, aord_num x = x * kappa_inv asymp_jasa_pderiv(p+1, i) = x end do end do end function qmckl_compute_jastrow_champ_asymp_jasa_pderiv
4.2.7. Parameter Derivative
Calculate the parameter derivatives of the electron-nucleus jastrow component factor_en_pderiv using the a_vector
coeffecients and the electron-nucleus rescaled distances
en_distance_rescaled. The result is stored in factor_en_pderiv. If
\(\alpha\) only runs over atoms of the same type then the
expressions for the derivatives are:
\[ \frac{\partial f}{\partial A_0} = \sum_{i, \alpha} \frac{
R_{i\alpha} }{1 + A_1 R_{i\alpha} } - \frac{\partial
J_{en}^\infty}{\partial A_0} \]
\[ \frac{\partial f}{\partial A_1} = \sum_{i, \alpha} \frac{
A_0 R_{i\alpha}^2 }{(1 + A_1 R_{i\alpha})^2 } - \frac{\partial
J_{en}^\infty}{\partial A_1} \]
\[ \frac{\partial f}{\partial A_k} = \sum_{i, \alpha} R_{i\alpha}^k - \frac{\partial
J_{en}^\infty}{\partial A_k}
\]
4.2.7.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_en_pderiv(qmckl_context context, double* const factor_en_pderiv, const int64_t size_max);
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_en_pderiv (context, & factor_en_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_en_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_en_pderiv end interface
4.2.7.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of unique nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | IDs of unique nuclei |
aord_num |
int64_t |
in | Number of coefficients |
a_vector |
double[type_nucl_num][aord_num+1] |
in | List of coefficients |
en_distance_rescaled |
double[walk_num][nucl_num][elec_num] |
in | Electron-nucleus distances |
asymp_jasa_pderiv |
double[type_nucl_num][aord_num+1] |
in | Type of nuclei |
factor_en_pderiv |
double[type_nucl_num][aord_num+1] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_en_pderiv_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, & type_nucl_vector, aord_num, a_vector, & en_distance_rescaled, asymp_jasa_pderiv, factor_en_pderiv) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: aord_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: en_distance_rescaled(elec_num,nucl_num,walk_num) real (c_double ) , intent(in) :: asymp_jasa_pderiv(aord_num+1,type_nucl_num) real (c_double ) , intent(out) :: factor_en_pderiv(aord_num+1,type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, p, nw double precision :: x, power_ser info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (type_nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_7 return endif factor_en_pderiv = 0.0d0 do nw =1, walk_num do a = 1, nucl_num do i = 1, elec_num x = en_distance_rescaled(i, a, nw) factor_en_pderiv(1,type_nucl_vector(a)+1) = factor_en_pderiv(1,type_nucl_vector(a)+1) + & x / (1.0d0 + a_vector(2, type_nucl_vector(a)+1) * x) - asymp_jasa_pderiv(1,type_nucl_vector(a)+1) factor_en_pderiv(2,type_nucl_vector(a)+1) = factor_en_pderiv(2,type_nucl_vector(a)+1) - & a_vector(2, type_nucl_vector(a)+1) * x**2 / (1.0d0 + a_vector(2, type_nucl_vector(a)+1) * x)**2 & - asymp_jasa_pderiv(2,type_nucl_vector(a)+1) do p = 2, aord_num x = x * en_distance_rescaled(i, a, nw) factor_en_pderiv(p+1,type_nucl_vector(a)+1) = factor_en_pderiv(p+1,type_nucl_vector(a)+1) + x & - asymp_jasa_pderiv(p+1,type_nucl_vector(a)+1) end do end do end do end do factor_en_pderiv = factor_en_pderiv / dble(walk_num) end function qmckl_compute_jastrow_champ_factor_en_pderiv_doc
4.2.8. Parameter Derivative of the gradient and Laplacian
The derivative of factor_en_gl with respect to all the jastrow parameters C is computed and stored into factor_en_gl_pderiv
4.2.8.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_en_gl_pderiv(qmckl_context context, double* const factor_en_gl_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_en_gl_pderiv (context, & factor_en_gl_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_en_gl_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_en_gl_pderiv end interface
4.2.8.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of unique nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | IDs of unique nuclei |
aord_num |
int64_t |
in | Number of coefficients |
a_vector |
double[type_nucl_num][aord_num+1] |
in | List of coefficients |
en_distance_rescaled |
double[walk_num][nucl_num][elec_num] |
in | Electron-nucleus distances |
en_distance_rescaled_gl |
double[walk_num][nucl_num][elec_num][4] |
in | Electron-nucleus distance derivatives |
factor_en_gl_pderiv |
double[type_nucl_num][aord_num+1][elec_num][4] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_en_gl_pderiv_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, & type_nucl_vector, aord_num, a_vector, & en_distance_rescaled, en_distance_rescaled_gl, factor_en_gl_pderiv) & bind(C) result(info) use qmckl implicit none integer (qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: aord_num real (c_double ) , intent(in) :: a_vector(aord_num+1,type_nucl_num) real (c_double ) , intent(in) :: en_distance_rescaled(elec_num,nucl_num,walk_num) real (c_double ) , intent(in) :: en_distance_rescaled_gl(4, elec_num,nucl_num,walk_num) real (c_double ) , intent(out) :: factor_en_gl_pderiv(4,elec_num,aord_num+1,type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, k, nw, ii double precision :: x, x1, kf double precision :: denom, invdenom, invdenom2, invdenom3, f double precision :: grad_c2 double precision :: dx(4) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (type_nucl_num <= 0) then info = QMCKL_INVALID_ARG_5 return endif if (aord_num < 0) then info = QMCKL_INVALID_ARG_7 return endif factor_en_gl_pderiv = 0.0d0 do nw =1, walk_num do a = 1, nucl_num do i = 1, elec_num x = en_distance_rescaled(i,a,nw) if(abs(x) < 1.d-12) continue denom = 1.0d0 + a_vector(2, type_nucl_vector(a)+1) * x invdenom = 1.0d0 / denom invdenom2 = invdenom*invdenom invdenom3 = invdenom2*invdenom dx(1) = en_distance_rescaled_gl(1,i,a,nw) dx(2) = en_distance_rescaled_gl(2,i,a,nw) dx(3) = en_distance_rescaled_gl(3,i,a,nw) dx(4) = en_distance_rescaled_gl(4,i,a,nw) f = invdenom2 grad_c2 = dx(1)*dx(1) + dx(2)*dx(2) + dx(3)*dx(3) factor_en_gl_pderiv(1,i,1,type_nucl_vector(a)+1) = factor_en_gl_pderiv(1,i,1,type_nucl_vector(a)+1) + f * dx(1) factor_en_gl_pderiv(2,i,1,type_nucl_vector(a)+1) = factor_en_gl_pderiv(2,i,1,type_nucl_vector(a)+1) + f * dx(2) factor_en_gl_pderiv(3,i,1,type_nucl_vector(a)+1) = factor_en_gl_pderiv(3,i,1,type_nucl_vector(a)+1) + f * dx(3) factor_en_gl_pderiv(4,i,1,type_nucl_vector(a)+1) = factor_en_gl_pderiv(4,i,1,type_nucl_vector(a)+1) & + f * (dx(4) - 2.d0 * a_vector(2, type_nucl_vector(a)+1) * grad_c2 * invdenom) f = -2.d0 * a_vector(1, type_nucl_vector(a)+1) * invdenom3 factor_en_gl_pderiv(1,i,2,type_nucl_vector(a)+1) = factor_en_gl_pderiv(1,i,2,type_nucl_vector(a)+1) + f * x * dx(1) factor_en_gl_pderiv(2,i,2,type_nucl_vector(a)+1) = factor_en_gl_pderiv(2,i,2,type_nucl_vector(a)+1) + f * x * dx(2) factor_en_gl_pderiv(3,i,2,type_nucl_vector(a)+1) = factor_en_gl_pderiv(3,i,2,type_nucl_vector(a)+1) + f * x * dx(3) factor_en_gl_pderiv(4,i,2,type_nucl_vector(a)+1) = factor_en_gl_pderiv(4,i,2,type_nucl_vector(a)+1) & + f * (x * dx(4) + invdenom * (grad_c2 - 2 * grad_c2 * x * a_vector(2, type_nucl_vector(a)+1))) kf = 2.d0 x1 = x x = 1.d0 do k=3, aord_num+1 f = kf * x factor_en_gl_pderiv(1,i,k,type_nucl_vector(a)+1) = factor_en_gl_pderiv(1,i,k,type_nucl_vector(a)+1) + f * x1 * dx(1) factor_en_gl_pderiv(2,i,k,type_nucl_vector(a)+1) = factor_en_gl_pderiv(2,i,k,type_nucl_vector(a)+1) + f * x1 * dx(2) factor_en_gl_pderiv(3,i,k,type_nucl_vector(a)+1) = factor_en_gl_pderiv(3,i,k,type_nucl_vector(a)+1) + f * x1 * dx(3) factor_en_gl_pderiv(4,i,k,type_nucl_vector(a)+1) = factor_en_gl_pderiv(4,i,k,type_nucl_vector(a)+1) & + f * (x1 * dx(4) + (kf-1.d0) * grad_c2) x = x*x1 kf = kf + 1.d0 end do end do end do end do factor_en_gl_pderiv = factor_en_gl_pderiv / dble(walk_num) end function qmckl_compute_jastrow_champ_factor_en_gl_pderiv_doc
4.3. Electron-electron-nucleus component
4.3.1. Electron-electron rescaled distances in \(J_\text{eeN}\)
een_rescaled_e stores the table of the rescaled distances between all
pairs of electrons and raised to the power \(p\) defined by cord_num:
\[ [g_e(r)_{ij}]^p = \begin{cases} \exp\left(-p\,\kappa_\text{e}\, r_{ij}\right) & \text{for } i \ne j \\ 0 & \text{for } i = j \]
where \(r_{ij}\) is the matrix of electron-electron distances.
4.3.1.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_een_rescaled_e(qmckl_context context, double* const een_rescaled_e, const int64_t size_max);
4.3.1.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
cord_num |
int64_t |
in | Order of polynomials |
rescale_factor_ee |
double |
in | Factor to rescale ee distances |
ee_distance |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances for each walker |
een_rescaled_e |
double[walk_num][0:cord_num][elec_num][elec_num] |
out | Electron-electron rescaled distances for each walker |
function qmckl_compute_een_rescaled_e_doc( & context, walk_num, elec_num, cord_num, rescale_factor_ee, & ee_distance, een_rescaled_e) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: cord_num real (c_double ) , intent(in) , value :: rescale_factor_ee real (c_double ) , intent(in) :: ee_distance(elec_num,elec_num,walk_num) real (c_double ) , intent(out) :: een_rescaled_e(elec_num,elec_num,0:cord_num,walk_num) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, j, k, l, nw info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (cord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif do nw = 1, walk_num do l = 0, cord_num do j = 1, elec_num do i = 1, elec_num een_rescaled_e(i, j, l, nw) = dexp(-rescale_factor_ee * ee_distance(i, j, nw))**l end do een_rescaled_e(j, j, l, nw) = 0.d0 end do end do end do end function qmckl_compute_een_rescaled_e_doc
4.3.1.3. Test
4.3.2. Electron-electron rescaled distances derivatives in \(J_\text{eeN}\)
een_rescaled_e_gl stores the table of the derivatives of the
rescaled distances between all pairs of electrons and raised to the
power \(p\) defined by cord_num. Here we take its derivatives
required for the een jastrowchamp.
\[ \frac{\partial}{\partial x} \left[ {g_\text{e}(r)}\right]^p = -\frac{x}{r} \kappa_\text{e}\, p\,\left[ {g_\text{e}(r)}\right]^p \]
\[ \Delta \left[ {g_\text{e}(r)}\right]^p = \kappa_\text{e}\, p\,\left[ - \frac{2}{r} + \kappa_\text{e}\, p \right] \left[ {g_\text{e}(r)} \right]^p \]
Derivatives are set to zero at \(r_{ii}\) to avoid NaNs.
4.3.2.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_een_rescaled_e_gl(qmckl_context context, double* const een_rescaled_e_gl, const int64_t size_max);
4.3.2.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
cord_num |
int64_t |
in | Order of polynomials |
rescale_factor_ee |
double |
in | Factor to rescale ee distances |
coord_ee |
double[3][walk_num][elec_num] |
in | Electron coordinates |
ee_distance |
double[walk_num][elec_num][elec_num] |
in | Electron-electron distances |
een_rescaled_e |
double[walk_num][0:cord_num][elec_num][elec_num] |
in | Electron-electron distances |
een_rescaled_e_gl |
double[walk_num][0:cord_num][elec_num][4][elec_num] |
out | Electron-electron rescaled distances |
function qmckl_compute_jastrow_champ_factor_een_rescaled_e_gl_doc( & context, walk_num, elec_num, cord_num, rescale_factor_ee, & coord_ee, ee_distance, een_rescaled_e, een_rescaled_e_gl) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer(c_int64_t) , intent(in), value :: walk_num integer(c_int64_t) , intent(in), value :: elec_num integer(c_int64_t) , intent(in), value :: cord_num real(c_double) , intent(in), value :: rescale_factor_ee real(c_double) , intent(in) :: coord_ee(elec_num,walk_num,3) real(c_double) , intent(in) :: ee_distance(elec_num,elec_num,walk_num) real(c_double) , intent(in) :: een_rescaled_e(elec_num,elec_num,0:cord_num,walk_num) real(c_double) , intent(out) :: een_rescaled_e_gl(elec_num,4,elec_num,0:cord_num,walk_num) integer(qmckl_exit_code) :: info double precision, allocatable :: elec_dist_gl(:,:,:) double precision :: x, kappa_l integer*8 :: i, j, k, l, nw, ii double precision :: rij_inv(elec_num) allocate(elec_dist_gl(elec_num, 4, elec_num)) elec_dist_gl = 0.d0 een_rescaled_e_gl = 0.d0 info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (cord_num < 0) then info = QMCKL_INVALID_ARG_4 return endif do nw = 1, walk_num ! Prepare table of exponentiated distances raised to appropriate power do j = 1, elec_num do i = 1, j-1 rij_inv(i) = 1.0d0 / ee_distance(i, j, nw) enddo rij_inv(j) = 0.0d0 do i = j+1, elec_num rij_inv(i) = 1.0d0 / ee_distance(i, j, nw) enddo do i = 1, elec_num do ii = 1, 3 elec_dist_gl(i, ii, j) = (coord_ee(i, nw, ii) - coord_ee(j, nw, ii)) * rij_inv(i) end do elec_dist_gl(i, 4, j) = 2.0d0 * rij_inv(i) end do end do ! Not necessary: should be set to zero by qmckl_malloc een_rescaled_e_gl(:,:,:,0,nw) = 0.d0 do l = 1, cord_num kappa_l = -dble(l) * rescale_factor_ee do j = 1, elec_num do i = 1, elec_num if (i /= j) then een_rescaled_e_gl(i, 1, j, l, nw) = kappa_l * elec_dist_gl(i, 1, j) * een_rescaled_e(i,j,l,nw) een_rescaled_e_gl(i, 2, j, l, nw) = kappa_l * elec_dist_gl(i, 2, j) * een_rescaled_e(i,j,l,nw) een_rescaled_e_gl(i, 3, j, l, nw) = kappa_l * elec_dist_gl(i, 3, j) * een_rescaled_e(i,j,l,nw) een_rescaled_e_gl(i, 4, j, l, nw) = kappa_l * (elec_dist_gl(i, 4, j) + kappa_l) * een_rescaled_e(i,j,l,nw) else een_rescaled_e_gl(i, 1, j, l, nw) = 0.d0 een_rescaled_e_gl(i, 2, j, l, nw) = 0.d0 een_rescaled_e_gl(i, 3, j, l, nw) = 0.d0 een_rescaled_e_gl(i, 4, j, l, nw) = 0.d0 end if end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_rescaled_e_gl_doc
4.3.3. Electron-nucleus rescaled distances in \(J_\text{eeN}\)
een_rescaled_n stores the table of the rescaled distances between
electrons and nuclei raised to the power \(p\) defined by cord_num:
\[ [g_{\alpha}(R_{i\alpha})]^p = \exp\left(-p\, \kappa_\alpha\, R_{i\alpha}\right) \]
where \(R_{i\alpha}\) is the matrix of electron-nucleus distances.
4.3.3.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_een_rescaled_n(qmckl_context context, double* const een_rescaled_n, const int64_t size_max);
4.3.3.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of atoms |
type_nucl_num |
int64_t |
in | Number of atom types |
type_nucl_vector |
int64_t[nucl_num] |
in | Types of atoms |
cord_num |
int64_t |
in | Order of polynomials |
rescale_factor_en |
double[nucl_num] |
in | Factor to rescale ee distances |
en_distance |
double[walk_num][elec_num][nucl_num] |
in | Electron-nucleus distances |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
out | Electron-nucleus rescaled distances |
function qmckl_compute_een_rescaled_n( & context, walk_num, elec_num, nucl_num, & type_nucl_num, type_nucl_vector, cord_num, rescale_factor_en, & en_distance, een_rescaled_n) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: cord_num real (c_double ) , intent(in) :: rescale_factor_en(nucl_num) real (c_double ) , intent(in) :: en_distance(nucl_num,elec_num,walk_num) real (c_double ) , intent(out) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, a, k, l, nw info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (cord_num < 0) then info = QMCKL_INVALID_ARG_5 return endif do nw = 1, walk_num ! prepare the actual een table een_rescaled_n(:, :, 0, nw) = 1.0d0 do a = 1, nucl_num do i = 1, elec_num een_rescaled_n(i, a, 1, nw) = dexp(-rescale_factor_en(type_nucl_vector(a)+1) * en_distance(a, i, nw)) end do end do do l = 2, cord_num do a = 1, nucl_num do i = 1, elec_num een_rescaled_n(i, a, l, nw) = een_rescaled_n(i, a, l - 1, nw) * een_rescaled_n(i, a, 1, nw) end do end do end do end do end function qmckl_compute_een_rescaled_n
4.3.4. Electron-nucleus rescaled distances derivatives in \(J_\text{eeN}\)
een_rescaled_n_gl stores the table of the derivatives of the
rescaled distances between all electron-nucleus pairs and raised to the
power \(p\) defined by cord_num. Here we take its derivatives
required for the een jastrowchamp.
\[ \frac{\partial}{\partial x} \left[ {g_\alpha(R_{i\alpha})}\right]^p = -\frac{x}{R_{i\alpha}} \kappa_\alpha\, p\,\left[ {g_\alpha(R_{i\alpha})}\right]^p \] \[ \Delta \left[ {g_\alpha(R_{i\alpha})}\right]^p = \frac{2}{R_{i\alpha}} \kappa_\alpha\, p\,\left[ {g_\alpha(R_{i\alpha})}\right]^p \right] + \left(\frac{\partial}{\partial x}\left[ {g_\alpha(R_{i\alpha})}\right]^p \right)^2 + \left(\frac{\partial}{\partial y}\left[ {g_\alpha(R_{i\alpha})}\right]^p \right)^2 + \left(\frac{\partial}{\partial z}\left[ {g_\alpha(R_{i\alpha})}\right]^p \right)^2 \]
4.3.4.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_een_rescaled_n_gl(qmckl_context context, double* const een_rescaled_n_gl, const int64_t size_max);
4.3.4.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of atoms |
type_nucl_num |
int64_t |
in | Number of atom types |
type_nucl_vector |
int64_t[nucl_num] |
in | Types of atoms |
cord_num |
int64_t |
in | Order of polynomials |
rescale_factor_en |
double[nucl_num] |
in | Factor to rescale ee distances |
coord_ee |
double[3][walk_num][elec_num] |
in | Electron coordinates |
coord_n |
double[3][nucl_num] |
in | Nuclear coordinates |
en_distance |
double[walk_num][elec_num][nucl_num] |
in | Electron-nucleus distances |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus distances |
een_rescaled_n_gl |
double[walk_num][0:cord_num][nucl_num][4][elec_num] |
out | Electron-nucleus rescaled distances |
function qmckl_compute_jastrow_champ_factor_een_rescaled_n_gl( & context, walk_num, elec_num, nucl_num, type_nucl_num, type_nucl_vector, & cord_num, rescale_factor_en, & coord_ee, coord_n, en_distance, een_rescaled_n, een_rescaled_n_gl) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: cord_num real (c_double ) , intent(in) :: rescale_factor_en(nucl_num) real (c_double ) , intent(in) :: coord_ee(elec_num,walk_num,3) real (c_double ) , intent(in) :: coord_n(nucl_num,3) real (c_double ) , intent(in) :: en_distance(nucl_num,elec_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: een_rescaled_n_gl(elec_num,4,nucl_num,0:cord_num,walk_num) integer(qmckl_exit_code) :: info double precision,allocatable :: elnuc_dist_gl(:,:,:) double precision :: x, ria_inv, kappa_l integer*8 :: i, a, k, l, nw, ii allocate(elnuc_dist_gl(elec_num, 4, nucl_num)) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif if (elec_num <= 0) then info = QMCKL_INVALID_ARG_3 return endif if (nucl_num <= 0) then info = QMCKL_INVALID_ARG_4 return endif if (cord_num < 0) then info = QMCKL_INVALID_ARG_5 return endif ! Prepare table of exponentiated distances raised to appropriate power een_rescaled_n_gl = 0.0d0 do nw = 1, walk_num ! prepare the actual een table do a = 1, nucl_num do i = 1, elec_num ria_inv = 1.0d0 / en_distance(a, i, nw) do ii = 1, 3 elnuc_dist_gl(i, ii, a) = (coord_ee(i, nw, ii) - coord_n(a, ii)) * ria_inv end do elnuc_dist_gl(i, 4, a) = 2.0d0 * ria_inv end do end do do l = 0, cord_num do a = 1, nucl_num kappa_l = - dble(l) * rescale_factor_en(type_nucl_vector(a)+1) do i = 1, elec_num een_rescaled_n_gl(i, 1, a, l, nw) = kappa_l * elnuc_dist_gl(i, 1, a) een_rescaled_n_gl(i, 2, a, l, nw) = kappa_l * elnuc_dist_gl(i, 2, a) een_rescaled_n_gl(i, 3, a, l, nw) = kappa_l * elnuc_dist_gl(i, 3, a) een_rescaled_n_gl(i, 4, a, l, nw) = kappa_l * elnuc_dist_gl(i, 4, a) een_rescaled_n_gl(i, 4, a, l, nw) = een_rescaled_n_gl(i, 4, a, l, nw) & + een_rescaled_n_gl(i, 1, a, l, nw) * een_rescaled_n_gl(i, 1, a, l, nw) & + een_rescaled_n_gl(i, 2, a, l, nw) * een_rescaled_n_gl(i, 2, a, l, nw) & + een_rescaled_n_gl(i, 3, a, l, nw) * een_rescaled_n_gl(i, 3, a, l, nw) een_rescaled_n_gl(i, 1, a, l, nw) = een_rescaled_n_gl(i, 1, a, l, nw) * & een_rescaled_n(i, a, l, nw) een_rescaled_n_gl(i, 2, a, l, nw) = een_rescaled_n_gl(i, 2, a, l, nw) * & een_rescaled_n(i, a, l, nw) een_rescaled_n_gl(i, 3, a, l, nw) = een_rescaled_n_gl(i, 3, a, l, nw) * & een_rescaled_n(i, a, l, nw) een_rescaled_n_gl(i, 4, a, l, nw) = een_rescaled_n_gl(i, 4, a, l, nw) * & een_rescaled_n(i, a, l, nw) end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_rescaled_n_gl
4.3.5. Temporary arrays for electron-electron-nucleus Jastrow \(f_{een}\)
Prepare c_vector_full and lkpm_combined_index tables required for the
calculation of the three-body jastrow factor_een and its derivative
factor_een_gl.
The array tmp_c corresponds to the tensor \(P\) defined at the
beginning of this section:
\[ P_{i\alpha}^{km} = \sum_{j=1}^{N_{\text{elec}}} \left[ g_\text{e}({r}_{ij}) \right]^k \left[ g_\alpha({R}_{j\alpha}) \right]^{m} \]
\[ \nabla_i P_{i\alpha}^{km} = \sum_{j=1}^{N_{\text{elec}}} \nabla_i \left[ g_\text{e}({r}_{ij}) \right]^k \left[ g_\alpha({R}_{j\alpha}) \right]^{m} \]
4.3.5.1. Compute dimcvector
Computes the dimension of the vector of parameters.
| \(N_{ord}\) | Number of parameters |
| 1 | 0 |
| 2 | 2 |
| 3 | 6 |
| 4 | 13 |
| 5 | 23 |
| 6 | 37 |
| 7 | 55 |
| 8 | 78 |
| 9 | 106 |
| 10 | 140 |
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
cord_num |
int64_t |
in | Order of polynomials |
dim_c_vector |
int64_t |
out | Number of parameters per atom type |
function qmckl_compute_dim_c_vector_doc( & context, cord_num, dim_c_vector) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(out) :: dim_c_vector integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, a, k, l, p, lmax info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (cord_num < 0) then info = QMCKL_INVALID_ARG_2 return endif dim_c_vector = 0 do p = 2, cord_num do k = p - 1, 0, -1 if (k .ne. 0) then lmax = p - k else lmax = p - k - 2 endif do l = lmax, 0, -1 if (iand(p - k - l, 1_8) == 1) cycle dim_c_vector = dim_c_vector + 1 end do end do end do end function qmckl_compute_dim_c_vector_doc
4.3.5.2. Get
4.3.5.3. Compute cvectorfull
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
nucl_num |
int64_t |
in | Number of atoms |
dim_c_vector |
int64_t |
in | dimension of cord full table |
type_nucl_num |
int64_t |
in | dimension of cord full table |
type_nucl_vector |
int64_t[nucl_num] |
in | dimension of cord full table |
c_vector |
double[dim_c_vector][type_nucl_num] |
in | dimension of cord full table |
c_vector_full |
double[dim_c_vector][nucl_num] |
out | Full list of coefficients |
function qmckl_compute_c_vector_full_doc( & context, nucl_num, dim_c_vector, type_nucl_num, & type_nucl_vector, c_vector, c_vector_full) & result(info) bind(C) 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 integer (c_int64_t) , intent(in) , value :: dim_c_vector integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) real (c_double ) , intent(in) :: c_vector(dim_c_vector, type_nucl_num) real (c_double ) , intent(out) :: c_vector_full(nucl_num,dim_c_vector) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, a, k, l, nw info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (nucl_num <= 0) info = QMCKL_INVALID_ARG_2 if (dim_c_vector < 0) info = QMCKL_INVALID_ARG_3 if (type_nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (info /= QMCKL_SUCCESS) return do a = 1, nucl_num c_vector_full(a,1:dim_c_vector) = c_vector(1:dim_c_vector, type_nucl_vector(a)+1) end do end function qmckl_compute_c_vector_full_doc
4.3.5.4. Compute lkpmcombinedindex
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
cord_num |
int64_t |
in | Order of polynomials |
dim_c_vector |
int64_t |
in | dimension of cord full table |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
out | Full list of combined indices |
function qmckl_compute_lkpm_combined_index_doc( & context, cord_num, dim_c_vector, lkpm_combined_index) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector integer (c_int64_t) , intent(out) :: lkpm_combined_index(dim_c_vector,4) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, a, k, l, kk, p, lmax, m info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (cord_num < 0) info = QMCKL_INVALID_ARG_2 if (dim_c_vector < 0) info = QMCKL_INVALID_ARG_3 if (info /= QMCKL_SUCCESS) return kk = 0 do p = 2, cord_num do k = p - 1, 0, -1 if (k /= 0) then lmax = p - k else lmax = p - k - 2 end if do l = lmax, 0, -1 if (iand(p - k - l, 1_8) .eq. 1_8) cycle m = (p - k - l)/2 kk = kk + 1 lkpm_combined_index(kk, 1) = l lkpm_combined_index(kk, 2) = k lkpm_combined_index(kk, 3) = p lkpm_combined_index(kk, 4) = m end do end do end do end function qmckl_compute_lkpm_combined_index_doc
4.3.5.5. Compute tmpc
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
cord_num |
int64_t |
in | Order of polynomials |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
walk_num |
int64_t |
in | Number of walkers |
een_rescaled_e |
double[walk_num][0:cord_num][elec_num][elec_num] |
in | Electron-electron rescaled factor |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
out | vector of non-zero coefficients |
function qmckl_compute_tmp_c_doc( & context, cord_num, elec_num, nucl_num, & walk_num, een_rescaled_e, een_rescaled_n, tmp_c) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: een_rescaled_e(elec_num,elec_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, j, a, l, kk, p, lmax, nw character :: TransA, TransB double precision :: alpha, beta integer*8 :: M, N, K, LDA, LDB, LDC TransA = 'N' TransB = 'N' alpha = 1.0d0 beta = 0.0d0 info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (cord_num < 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (walk_num <= 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return M = elec_num K = elec_num N = nucl_num LDA = size(een_rescaled_e,1) LDB = size(een_rescaled_n,1) LDC = size(tmp_c,1) do nw=1, walk_num do i=0, cord_num-1 do j=0, cord_num-i info = qmckl_dgemm(context, TransA, TransB, M, N, K, alpha, & een_rescaled_e(1,1,i,nw),LDA*1_8, & een_rescaled_n(1,1,j,nw),LDB*1_8, & beta, & tmp_c(1,1,j,i,nw),LDC) end do end do end do end function qmckl_compute_tmp_c_doc
4.3.5.6. Compute dtmpc
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
cord_num |
int64_t |
in | Order of polynomials |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
walk_num |
int64_t |
in | Number of walkers |
een_rescaled_e_gl |
double[walk_num][0:cord_num][elec_num][4][elec_num] |
in | Electron-electron rescaled factor derivatives |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
dtmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][4][elec_num] |
out | vector of non-zero coefficients |
function qmckl_compute_dtmp_c_doc( & context, cord_num, elec_num, nucl_num, & walk_num, een_rescaled_e_gl, een_rescaled_n, dtmp_c) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: een_rescaled_e_gl(elec_num,4,elec_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: dtmp_c(elec_num,4,nucl_num,0:cord_num,0:cord_num-1,walk_num) integer(qmckl_exit_code) :: info double precision :: x integer*8 :: i, j, a, l, kk, p, lmax, nw, ii character :: TransA, TransB double precision :: alpha, beta integer*8 :: M, N, K, LDA, LDB, LDC info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (cord_num < 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (walk_num <= 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return TransA = 'N' TransB = 'N' alpha = 1.0d0 beta = 0.0d0 M = 4*elec_num N = nucl_num K = elec_num LDA = 4*size(een_rescaled_e_gl,1) LDB = size(een_rescaled_n,1) LDC = 4*size(dtmp_c,1) do nw=1, walk_num do i=0, cord_num-1 do j=0, cord_num-i info = qmckl_dgemm(context,TransA, TransB, M, N, K, alpha, & een_rescaled_e_gl(1,1,1,i,nw),LDA*1_8, & een_rescaled_n(1,1,j,nw),LDB*1_8, & beta, & dtmp_c(1,1,1,j,i,nw),LDC) end do end do end do end function qmckl_compute_dtmp_c_doc
4.3.6. Electron-electron-nucleus Jastrow \(f_{een}\)
Calculate the electron-electron-nuclear three-body jastrow component factor_een
using the above prepared tables.
TODO: write equations.
4.3.6.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_een(qmckl_context context, double* const factor_een, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_een (context, & factor_een, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_een(size_max) end function qmckl_get_jastrow_champ_factor_een end interface
4.3.6.2. Compute naive
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
een_rescaled_e |
double[walk_num][0:cord_num][elec_num][elec_num] |
in | Electron-nucleus rescaled |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
factor_een |
double[walk_num] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_naive( & context, walk_num, elec_num, nucl_num, cord_num,& dim_c_vector, c_vector_full, lkpm_combined_index, & een_rescaled_e, een_rescaled_n, factor_een) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: een_rescaled_e(elec_num,elec_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een(walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, m, n, p, nw double precision :: cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return do nw =1, walk_num factor_een(nw) = 0.d0 do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) p = lkpm_combined_index(n, 3) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) if (cn == 0.d0) cycle do j = 1, elec_num do i = 1, j-1 factor_een(nw) = factor_een(nw) + cn*( & een_rescaled_e(i,j,k,nw) * & (een_rescaled_n(i,a,l,nw) + een_rescaled_n(j,a,l,nw)) * & (een_rescaled_n(i,a,m,nw) * een_rescaled_n(j,a,m,nw)) ) end do end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_naive
4.3.6.3. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
in | vector of non-zero coefficients |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled distances |
factor_een |
double[walk_num] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_doc( & context, walk_num, elec_num, nucl_num, cord_num, & dim_c_vector, c_vector_full, lkpm_combined_index, & tmp_c, een_rescaled_n, factor_een) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een(walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, p, m, n, nw double precision :: accu, accu2, cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return factor_een = 0.0d0 if (cord_num == 0) return do nw =1, walk_num do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) if(cn == 0.d0) cycle accu = 0.0d0 do j = 1, elec_num accu = accu + een_rescaled_n(j,a,m,nw) * tmp_c(j,a,m+l,k,nw) end do factor_een(nw) = factor_een(nw) + accu * cn end do end do end do end function qmckl_compute_jastrow_champ_factor_een_doc
4.3.7. Electron-electron-nucleus Jastrow \(f_{een}\) derivative
Calculate the electron-electron-nuclear three-body jastrow component factor_een_gl
using the above prepared tables.
TODO: write equations.
4.3.7.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_een_gl(qmckl_context context, double* const factor_een_gl, const int64_t size_max);
qmckl_exit_code qmckl_get_jastrow_champ_factor_een_grad(qmckl_context context, double* const factor_een_grad, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_een_gl (context, & factor_een_gl, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_een_gl(size_max) end function qmckl_get_jastrow_champ_factor_een_gl end interface interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_een_grad (context, & factor_een_grad, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_een_grad(size_max) end function qmckl_get_jastrow_champ_factor_een_grad end interface
4.3.7.2. Compute Naive
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
een_rescaled_e |
double[walk_num][0:cord_num][elec_num][elec_num] |
in | Electron-nucleus rescaled |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
een_rescaled_e_gl |
double[walk_num][0:cord_num][elec_num][4][elec_num] |
in | Electron-nucleus rescaled |
een_rescaled_n_gl |
double[walk_num][0:cord_num][nucl_num][4][elec_num] |
in | Electron-nucleus rescaled factor |
factor_een_gl |
double[walk_num][4][elec_num] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_gl_naive( & context, walk_num, elec_num, nucl_num, cord_num, dim_c_vector, & c_vector_full, lkpm_combined_index, een_rescaled_e, een_rescaled_n, & een_rescaled_e_gl, een_rescaled_n_gl, factor_een_gl)& result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: een_rescaled_e(elec_num,elec_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_e_gl(elec_num,4,elec_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n_gl(elec_num,4,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een_gl(elec_num,4,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, m, n, nw double precision :: accu, accu2, cn double precision :: daccu(1:4), daccu2(1:4) info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return factor_een_gl = 0.0d0 do nw =1, walk_num do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) do j = 1, elec_num accu = 0.0d0 accu2 = 0.0d0 daccu = 0.0d0 daccu2 = 0.0d0 do i = 1, elec_num accu = accu + een_rescaled_e(i, j, k, nw) * een_rescaled_n(i, a, m, nw) accu2 = accu2 + een_rescaled_e(i, j, k, nw) * een_rescaled_n(i, a, m + l, nw) daccu(1:4) = daccu(1:4) + een_rescaled_e_gl(j, 1:4, i, k, nw) * & een_rescaled_n(i, a, m, nw) daccu2(1:4) = daccu2(1:4) + een_rescaled_e_gl(j, 1:4, i, k, nw) * & een_rescaled_n(i, a, m + l, nw) end do factor_een_gl(j, 1:4, nw) = factor_een_gl(j, 1:4, nw) + & (accu * een_rescaled_n_gl(j, 1:4, a, m + l, nw) & + daccu(1:4) * een_rescaled_n(j, a, m + l, nw) & + daccu2(1:4) * een_rescaled_n(j, a, m, nw) & + accu2 * een_rescaled_n_gl(j, 1:4, a, m, nw)) * cn factor_een_gl(j, 4, nw) = factor_een_gl(j, 4, nw) + 2.0d0 * ( & daccu (1) * een_rescaled_n_gl(j, 1, a, m + l, nw) + & daccu (2) * een_rescaled_n_gl(j, 2, a, m + l, nw) + & daccu (3) * een_rescaled_n_gl(j, 3, a, m + l, nw) + & daccu2(1) * een_rescaled_n_gl(j, 1, a, m, nw ) + & daccu2(2) * een_rescaled_n_gl(j, 2, a, m, nw ) + & daccu2(3) * een_rescaled_n_gl(j, 3, a, m, nw ) ) * cn end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_gl_naive
4.3.7.3. Compute GL
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
in | Temporary intermediate tensor |
dtmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][4][elec_num] |
in | vector of non-zero coefficients |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
een_rescaled_n_gl |
double[walk_num][0:cord_num][nucl_num][4][elec_num] |
in | Derivative of Electron-nucleus rescaled factor |
factor_een_gl |
double[walk_num][4][elec_num] |
out | Derivative of Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_gl_doc( & context, walk_num, elec_num, nucl_num, & cord_num, dim_c_vector, c_vector_full, lkpm_combined_index, & tmp_c, dtmp_c, een_rescaled_n, een_rescaled_n_gl, factor_een_gl)& result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: dtmp_c(elec_num,4,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n_gl(elec_num,4,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een_gl(elec_num,4,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, m, n, nw, ii double precision :: accu, accu2, cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return if (cord_num == 0) then factor_een_gl = 0.0d0 return end if do nw =1, walk_num factor_een_gl(:,:,nw) = 0.0d0 do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) if(cn == 0.d0) cycle do ii = 1, 4 do j = 1, elec_num factor_een_gl(j,ii,nw) = factor_een_gl(j,ii,nw) + ( & tmp_c (j, a,m ,k,nw) * een_rescaled_n_gl(j,ii,a,m+l,nw) + & tmp_c (j, a,m+l,k,nw) * een_rescaled_n_gl(j,ii,a,m ,nw) + & dtmp_c(j,ii,a,m ,k,nw) * een_rescaled_n (j, a,m+l,nw) + & dtmp_c(j,ii,a,m+l,k,nw) * een_rescaled_n (j, a,m ,nw) & ) * cn end do end do cn = cn + cn do j = 1, elec_num factor_een_gl(j,4,nw) = factor_een_gl(j,4,nw) + ( & dtmp_c(j,1,a,m ,k,nw) * een_rescaled_n_gl(j,1,a,m+l,nw) + & dtmp_c(j,2,a,m ,k,nw) * een_rescaled_n_gl(j,2,a,m+l,nw) + & dtmp_c(j,3,a,m ,k,nw) * een_rescaled_n_gl(j,3,a,m+l,nw) + & dtmp_c(j,1,a,m+l,k,nw) * een_rescaled_n_gl(j,1,a,m ,nw) + & dtmp_c(j,2,a,m+l,k,nw) * een_rescaled_n_gl(j,2,a,m ,nw) + & dtmp_c(j,3,a,m+l,k,nw) * een_rescaled_n_gl(j,3,a,m ,nw) & ) * cn end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_gl_doc
4.3.7.4. Compute Gradient only
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
in | Temporary intermediate tensor |
dtmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][4][elec_num] |
in | vector of non-zero coefficients |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
een_rescaled_n_gl |
double[walk_num][0:cord_num][nucl_num][4][elec_num] |
in | Derivative of Electron-nucleus rescaled factor |
factor_een_grad |
double[walk_num][3][elec_num] |
out | Derivative of Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_grad_doc( & context, walk_num, elec_num, nucl_num, & cord_num, dim_c_vector, c_vector_full, lkpm_combined_index, & tmp_c, dtmp_c, een_rescaled_n, een_rescaled_n_gl, factor_een_grad) & 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 :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: dtmp_c(elec_num,4,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n_gl(elec_num,4,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een_grad(elec_num,3,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, m, n, nw, ii double precision :: accu, accu2, cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return if (cord_num == 0) then factor_een_grad = 0.0d0 return end if do nw =1, walk_num factor_een_grad(:,:,nw) = 0.0d0 do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) if(cn == 0.d0) cycle do ii = 1, 3 do j = 1, elec_num factor_een_grad(j,ii,nw) = factor_een_grad(j,ii,nw) + ( & dtmp_c(j,ii,a,m ,k,nw) * een_rescaled_n (j, a,m+l,nw) + & dtmp_c(j,ii,a,m+l,k,nw) * een_rescaled_n (j, a,m ,nw) + & tmp_c(j,a,m ,k,nw) * een_rescaled_n_gl(j,ii,a,m+l,nw) + & tmp_c(j,a,m+l,k,nw) * een_rescaled_n_gl(j,ii,a,m ,nw) & ) * cn end do end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_grad_doc
4.3.8. Electron-electron-nucleus Jastrow Parameter derivatives
Calculate the derivatives of electron-electron-nuclear three-body jastrow component factor_een
wit respect to the parameters.
TODO: write equations.
4.3.8.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_een_pderiv(qmckl_context context, double* const factor_een_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_een_pderiv (context, & factor_een_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_een_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_een_pderiv end interface
4.3.8.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of unique nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | IDs of unique nuclei |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
in | vector of non-zero coefficients |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled distances |
factor_een_pderiv |
double[type_nucl_num][dim_c_vector] |
out | Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_pderiv_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, & type_nucl_vector, cord_num, dim_c_vector, c_vector_full, & lkpm_combined_index, tmp_c, een_rescaled_n, factor_een_pderiv) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een_pderiv(dim_c_vector, type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, p, m, n, nw double precision :: accu, accu2, cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return factor_een_pderiv = 0.0d0 if (cord_num == 0) return do nw =1, walk_num do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num accu = 0.0d0 do j = 1, elec_num accu = accu + een_rescaled_n(j,a,m,nw) * tmp_c(j,a,m+l,k,nw) end do factor_een_pderiv(n, type_nucl_vector(a)+1) = factor_een_pderiv(n, type_nucl_vector(a)+1) + accu end do end do end do end function qmckl_compute_jastrow_champ_factor_een_pderiv_doc
4.3.9. Electron-electron-nucleus Parameter Derivative of the gradient and Laplacian
The derivative of factor_een_gl with respect to all the jastrow parameters
C is computed and stored into factor_een_gl_pderiv
TODO: write equations
4.3.9.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_factor_een_gl_pderiv(qmckl_context context, double* const factor_een_gl_pderiv, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_factor_een_gl_pderiv (context, & factor_een_gl_pderiv, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: factor_een_gl_pderiv(size_max) end function qmckl_get_jastrow_champ_factor_een_gl_pderiv end interface
4.3.9.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
nucl_num |
int64_t |
in | Number of nuclei |
type_nucl_num |
int64_t |
in | Number of types of nuclei |
type_nucl_vector |
int64_t[nucl_num] |
in | The nucl type index of each nucleus |
cord_num |
int64_t |
in | order of polynomials |
dim_c_vector |
int64_t |
in | dimension of full coefficient vector |
c_vector_full |
double[dim_c_vector][nucl_num] |
in | full coefficient vector |
lkpm_combined_index |
int64_t[4][dim_c_vector] |
in | combined indices |
tmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][elec_num] |
in | Temporary intermediate tensor |
dtmp_c |
double[walk_num][0:cord_num-1][0:cord_num][nucl_num][4][elec_num] |
in | vector of non-zero coefficients |
een_rescaled_n |
double[walk_num][0:cord_num][nucl_num][elec_num] |
in | Electron-nucleus rescaled factor |
een_rescaled_n_gl |
double[walk_num][0:cord_num][nucl_num][4][elec_num] |
in | Derivative of Electron-nucleus rescaled factor |
factor_een_gl_pderiv |
double[type_nucl_num][dim_c_vector][elec_num][4] |
out | Derivative of Electron-nucleus jastrow |
function qmckl_compute_jastrow_champ_factor_een_gl_pderiv_doc( & context, walk_num, elec_num, nucl_num, type_nucl_num, type_nucl_vector, & cord_num, dim_c_vector, c_vector_full, lkpm_combined_index, & tmp_c, dtmp_c, een_rescaled_n, een_rescaled_n_gl, factor_een_gl_pderiv)& result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num integer (c_int64_t) , intent(in) , value :: nucl_num integer (c_int64_t) , intent(in) , value :: type_nucl_num integer (c_int64_t) , intent(in) :: type_nucl_vector(nucl_num) integer (c_int64_t) , intent(in) , value :: cord_num integer (c_int64_t) , intent(in) , value :: dim_c_vector real (c_double ) , intent(in) :: c_vector_full(nucl_num,dim_c_vector) integer (c_int64_t) , intent(in) :: lkpm_combined_index(dim_c_vector,4) real (c_double ) , intent(in) :: tmp_c(elec_num,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: dtmp_c(elec_num,4,nucl_num,0:cord_num,0:cord_num-1,walk_num) real (c_double ) , intent(in) :: een_rescaled_n(elec_num,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(in) :: een_rescaled_n_gl(elec_num,4,nucl_num,0:cord_num,walk_num) real (c_double ) , intent(out) :: factor_een_gl_pderiv(4,elec_num,dim_c_vector,type_nucl_num) integer(qmckl_exit_code) :: info integer*8 :: i, a, j, l, k, m, n, nw, ii double precision :: accu, accu2, cn info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) info = QMCKL_INVALID_CONTEXT if (walk_num <= 0) info = QMCKL_INVALID_ARG_2 if (elec_num <= 0) info = QMCKL_INVALID_ARG_3 if (nucl_num <= 0) info = QMCKL_INVALID_ARG_4 if (cord_num < 0) info = QMCKL_INVALID_ARG_5 if (info /= QMCKL_SUCCESS) return factor_een_gl_pderiv = 0.0d0 if (cord_num == 0) then return end if do nw =1, walk_num do n = 1, dim_c_vector l = lkpm_combined_index(n, 1) k = lkpm_combined_index(n, 2) m = lkpm_combined_index(n, 4) do a = 1, nucl_num cn = c_vector_full(a, n) ! if(cn == 0.d0) cycle do ii = 1, 4 do j = 1, elec_num factor_een_gl_pderiv(ii,j,n,type_nucl_vector(a)+1) = factor_een_gl_pderiv(ii,j,n,type_nucl_vector(a)+1) + ( & tmp_c (j, a,m ,k,nw) * een_rescaled_n_gl(j,ii,a,m+l,nw) + & tmp_c (j, a,m+l,k,nw) * een_rescaled_n_gl(j,ii,a,m ,nw) + & dtmp_c(j,ii,a,m ,k,nw) * een_rescaled_n (j, a,m+l,nw) + & dtmp_c(j,ii,a,m+l,k,nw) * een_rescaled_n (j, a,m ,nw) & ) end do end do do j = 1, elec_num factor_een_gl_pderiv(4,j,n,type_nucl_vector(a)+1) = factor_een_gl_pderiv(4,j,n,type_nucl_vector(a)+1) + ( & dtmp_c(j,1,a,m ,k,nw) * een_rescaled_n_gl(j,1,a,m+l,nw) + & dtmp_c(j,2,a,m ,k,nw) * een_rescaled_n_gl(j,2,a,m+l,nw) + & dtmp_c(j,3,a,m ,k,nw) * een_rescaled_n_gl(j,3,a,m+l,nw) + & dtmp_c(j,1,a,m+l,k,nw) * een_rescaled_n_gl(j,1,a,m ,nw) + & dtmp_c(j,2,a,m+l,k,nw) * een_rescaled_n_gl(j,2,a,m ,nw) + & dtmp_c(j,3,a,m+l,k,nw) * een_rescaled_n_gl(j,3,a,m ,nw) & ) * 2 end do end do end do end do end function qmckl_compute_jastrow_champ_factor_een_gl_pderiv_doc
4.4. Total Jastrow
4.4.1. Value
Value of the total Jastrow factor: \(\exp(J)\)
4.4.1.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_value(qmckl_context context, double* const value, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_value (context, & value, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: value(size_max) end function qmckl_get_jastrow_champ_value end interface
4.4.1.2. Compute
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
f_ee |
double[walk_num] |
in | ee component |
f_en |
double[walk_num] |
in | eN component |
f_een |
double[walk_num] |
in | eeN component |
value |
double[walk_num] |
out | Total Jastrow factor |
function qmckl_compute_jastrow_champ_value_doc(context, & walk_num, f_ee, f_en, f_een, value) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num real (c_double ) , intent(in) :: f_ee(walk_num) real (c_double ) , intent(in) :: f_en(walk_num) real (c_double ) , intent(in) :: f_een(walk_num) real (c_double ) , intent(out) :: value(walk_num) integer(qmckl_exit_code) :: info integer*8 :: i info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif do i = 1, walk_num value(i) = f_ee(i) + f_en(i) + f_een(i) end do do i = 1, walk_num ! Flush to zero to avoid floating-point exception if (value(i) < -100.d0) then value(i) = 0.d0 else value(i) = dexp(value(i)) endif end do end function qmckl_compute_jastrow_champ_value_doc
4.4.2. Derivatives
Gradients and Laplacian of the total Jastrow factor: \[ \nabla \left[ e^{J(\mathbf{r})} \right] = e^{J(\mathbf{r})} \nabla J(\mathbf{r}) \] \[ \Delta \left[ e^{J(\mathbf{r})} \right] = e^{J(\mathbf{r})} \left[ \Delta J(\mathbf{r}) + \nabla J(\mathbf{r}) \cdot \nabla J(\mathbf{r}) \right] \]
4.4.2.1. Get
qmckl_exit_code qmckl_get_jastrow_champ_gl(qmckl_context context, double* const gl, const int64_t size_max); qmckl_exit_code qmckl_get_jastrow_champ_grad(qmckl_context context, double* const grad, const int64_t size_max);
qmckl_exit_code qmckl_get_jastrow_champ_grad(qmckl_context context, double* const grad, const int64_t size_max);
- Fortran interface
interface integer(qmckl_exit_code) function qmckl_get_jastrow_champ_gl (context, & gl, size_max) bind(C) use, intrinsic :: iso_c_binding import implicit none integer (qmckl_context) , intent(in), value :: context integer(c_int64_t), intent(in), value :: size_max real(c_double), intent(out) :: gl(size_max) end function qmckl_get_jastrow_champ_gl end interface
4.4.2.2. Compute GL
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
value |
double[walk_num] |
in | Total Jastrow |
gl_ee |
double[walk_num][4][elec_num] |
in | ee component |
gl_en |
double[walk_num][4][elec_num] |
in | eN component |
gl_een |
double[walk_num][4][elec_num] |
in | eeN component |
gl |
double[walk_num][4][elec_num] |
out | Total Jastrow factor |
function qmckl_compute_jastrow_champ_gl_doc(context, & walk_num, elec_num, value, gl_ee, gl_en, gl_een, gl) & bind(C) result(info) use qmckl use, intrinsic :: iso_c_binding implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num real (c_double ) , intent(in) :: value(walk_num) real (c_double ) , intent(in) :: gl_ee(elec_num,4,walk_num) real (c_double ) , intent(in) :: gl_en(elec_num,4,walk_num) real (c_double ) , intent(in) :: gl_een(elec_num,4,walk_num) real (c_double ) , intent(out) :: gl(elec_num,4,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, k info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif do k = 1, walk_num do j=1,4 do i = 1, elec_num gl(i,j,k) = gl_ee(i,j,k) + gl_en(i,j,k) + gl_een(i,j,k) end do end do do i = 1, elec_num gl(i,4,k) = gl(i,4,k) + & gl(i,1,k) * gl(i,1,k) + & gl(i,2,k) * gl(i,2,k) + & gl(i,3,k) * gl(i,3,k) end do gl(:,:,k) = gl(:,:,k) * value(k) end do end function qmckl_compute_jastrow_champ_gl_doc
4.4.2.3. Compute Gradient only
| Variable | Type | In/Out | Description |
|---|---|---|---|
context |
qmckl_context |
in | Global state |
walk_num |
int64_t |
in | Number of walkers |
elec_num |
int64_t |
in | Number of electrons |
value |
double[walk_num] |
in | Total Jastrow |
gl_ee |
double[walk_num][4][elec_num] |
in | ee component |
gl_en |
double[walk_num][4][elec_num] |
in | eN component |
grad_een |
double[walk_num][3][elec_num] |
in | eeN component |
grad |
double[walk_num][3][elec_num] |
out | Total Jastrow factor |
function qmckl_compute_jastrow_champ_grad_doc(context, & walk_num, elec_num, value, gl_ee, gl_en, grad_een, grad) & result(info) bind(C) use, intrinsic :: iso_c_binding use qmckl implicit none integer(qmckl_context), intent(in), value :: context integer (c_int64_t) , intent(in) , value :: walk_num integer (c_int64_t) , intent(in) , value :: elec_num real (c_double ) , intent(in) :: value(walk_num) real (c_double ) , intent(in) :: gl_ee(elec_num,4,walk_num) real (c_double ) , intent(in) :: gl_en(elec_num,4,walk_num) real (c_double ) , intent(in) :: grad_een(elec_num,3,walk_num) real (c_double ) , intent(out) :: grad(elec_num,3,walk_num) integer(qmckl_exit_code) :: info integer*8 :: i, j, k info = QMCKL_SUCCESS if (context == QMCKL_NULL_CONTEXT) then info = QMCKL_INVALID_CONTEXT return endif if (walk_num <= 0) then info = QMCKL_INVALID_ARG_2 return endif do k = 1, walk_num do j=1,3 do i = 1, elec_num grad(i,j,k) = gl_ee(i,j,k) + gl_en(i,j,k) + grad_een(i,j,k) end do end do grad(:,:,k) = grad(:,:,k) * value(k) end do end function qmckl_compute_jastrow_champ_grad_doc