Reference Input Parameter Library (RIPL)¶
The Reference Input Parameter Library (RIPL) contains nuclear structure information. These files are much easier to parse and manipulate than their source, the Experimental Nuclear Structure Database (XUNDL). XUNDL files are not directly supported for now. Let us get an overview of what NucML can do for us. Let us begin, as always, by importing the nucml.datasets module.
[1]:
# # PROTOTYPE
# import sys
# sys.path.append("../..")
[2]:
import nucml.datasets as nuc_data
RIPL/XUNDL Headers¶
The RIPL headers are the most basic information in terms of nuclear structure. It includes the elements, the number of levels and gammas available, and the neutron and proton separation energy.
[3]:
ensdf_headers = nuc_data.load_ensdf_headers()
[4]:
ensdf_headers.head()
[4]:
| Element_w_A | A | Z | Number_of_Levels | Number_of_Gammas | N_max | N_c | Sn | Sp | |
|---|---|---|---|---|---|---|---|---|---|
| 0 | 1H | 1 | 1 | 1 | 0 | 1 | 1 | 0.00000 | 0.00000 |
| 1 | 2H | 2 | 1 | 1 | 0 | 1 | 1 | 2.22457 | 2.22457 |
| 2 | 3H | 3 | 1 | 1 | 0 | 1 | 1 | 6.25723 | 0.00000 |
| 3 | 3He | 3 | 2 | 1 | 0 | 1 | 1 | 0.00000 | 5.49348 |
| 4 | 4He | 4 | 2 | 16 | 0 | 1 | 16 | 20.57762 | 19.81386 |
Elemental RIPL Information¶
We can easily load all available nuclear structure levels for a given isotope using the load_ensdf_isotopic(). It returns all levels and their energies, the spin, parity, half-life, gammas per level, a RIPL specific flag, and the number of decay modes.
[5]:
u236_levels = nuc_data.load_ensdf_isotopic("u236")
[6]:
u236_levels.head()
[6]:
| Level_Number | Energy | Spin | Parity | Half_Life | Gammas | Flag | ENSDF_Spin | Num_Decay_Modes | Decay_Info | |
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1.0 | 0.000000 | 0.0 | 1 | 7.391000e+14 | 0 | u | 0+ | 2 | = 1.0000E+02 %A = 9.4000E-08 %SF ... |
| 1 | 2.0 | 0.045244 | 2.0 | 1 | 2.340000e-10 | 1 | u | 2+ | 0 | 0.000000 0 |
| 2 | 3.0 | 0.149477 | 4.0 | 1 | 1.240000e-10 | 1 | u | 4+ | 0 | 0.000000 0 |
| 3 | 4.0 | 0.309785 | 6.0 | 1 | 5.800000e-11 | 1 | u | 6+ | 0 | 0.000000 0 |
| 4 | 5.0 | 0.522250 | 8.0 | 1 | 2.400000e-11 | 1 | u | 8+ | 0 | 0.000000 0 |
Elemental RIPL Gammas¶
We can also obtain the known gammas per level including their initial and final level state, the energy, the gamma and electromagnetic decay probability, and the ICC.
[7]:
u235_gammas = nuc_data.load_ensdf_isotopic("u235", filetype='gammas')
[8]:
u235_gammas.head()
[8]:
| Level_Record | Final_State | Energy | Gamma_Decay_Prob | Electromag_Decay_Prob | ICC | |
|---|---|---|---|---|---|---|
| 2 | 2.0 | 1 | 0.000 | 1.000000e-10 | 1.0000 | 1.000000e+10 |
| 4 | 3.0 | 2 | 0.013 | 9.990000e-04 | 1.0000 | 1.000000e+03 |
| 6 | 4.0 | 1 | 0.046 | 1.515000e-02 | 1.0000 | 6.500000e+01 |
| 8 | 5.0 | 3 | 0.039 | 8.712000e-04 | 0.2709 | 3.100000e+02 |
| 9 | 5.0 | 2 | 0.052 | 2.293000e-03 | 0.7291 | 3.170000e+02 |
Ground State Data¶
With the previous functions, you can obtain most of the basic nuclear structure data. Another commodity function offered by NucML is the load_ensdf_ground_states(). This will load only the ground state for all available isotopes.
[9]:
ground_states = nuc_data.load_ensdf_ground_states()
[10]:
ground_states.head()
[10]:
| Element_w_A | Spin | Parity | Half_Life | Flag | ENSDF_Spin | Num_Decay_Modes | Modifier | Decay_Info | |
|---|---|---|---|---|---|---|---|---|---|
| 0 | 1H | 0.5 | 1 | -1.0 | 1/2+ | 0 | 0.000000 | ||
| 1 | 2H | 1.0 | 1 | -1.0 | 1+ | 0 | 0.000000 | ||
| 2 | 3H | 0.5 | 1 | 388800000.0 | 1/2+ | 1 | = | 1.0000E+02 %B- ... | |
| 3 | 3He | 0.5 | 1 | -1.0 | 1/2+ | 0 | 0.000000 | ||
| 4 | 4He | 0.0 | 1 | -1.0 | 0+ | 0 | 0.000000 |
RIPL Cut-Off Parameters¶
RIPL offers some recommended cut-off parameters for nuclear structure data. These will be useful when deciding the transformations applied to an ML-ready XUNDL dataset. For now, you can obtain the Cut-Off parameters using the load_riple_parameteres().
[11]:
ripl_cutoff = nuc_data.load_ripl_parameters()
[12]:
ripl_cutoff.head()
[12]:
| Z | A | Element | Temperature_MeV | Temperature_U | Black_Shift | Black_Shift_U | N_Lev_ENSDF | N_Max_Lev_Complete | Min_Lev_Complete | Num_Lev_Unique_Spin | E_Max_N_Max | E_Num_Lev_U_Spin | Chi | Fit | Flag | Nox | Xm_Ex | Sigma | Element_w_A | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 0 | 1 | n | 0.0 | 0.0 | 0.0 | 0.0 | 1 | 1 | 1 | 1 | 0.0 | 0.0 | 0 | 0.0 | 1n | ||||
| 1 | 1 | 1 | H | 0.0 | 0.0 | 0.0 | 0.0 | 1 | 1 | 1 | 1 | 0.0 | 0.0 | 0 | 0.0 | 1H | ||||
| 2 | 1 | 2 | H | 0.0 | 0.0 | 0.0 | 0.0 | 1 | 1 | 1 | 1 | 0.0 | 0.0 | 0 | 0.0 | 2H | ||||
| 3 | 1 | 3 | H | 0.0 | 0.0 | 0.0 | 0.0 | 1 | 1 | 1 | 1 | 0.0 | 0.0 | 0 | 0.0 | 3H | ||||
| 4 | 2 | 3 | He | 0.0 | 0.0 | 0.0 | 0.0 | 1 | 1 | 1 | 1 | 0.0 | 0.0 | 0 | 0.0 | 3He |
All RIPL/XUNDL Data¶
You can also load all nuclear structure levels and their energies along with some basic information. It is a good place to get started building an ML-ready dataset.
[13]:
ensdf = nuc_data.load_ensdf()
INFO:root:Reading data from C:/Users/Pedro/Desktop/ML_Nuclear_Data/ENSDF\CSV_Files/ensdf.csv
[14]:
ensdf.head()
[14]:
| Level_Number | Energy | Spin | Parity | Half_Life | Gammas | Flag | ENSDF_Spin | Num_Decay_Modes | Decay_Info | Element_w_A | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 0.0 | 0.5 | 1 | -1.0 | 0 | NaN | 1/2+ | 0 | 0.0 | 1H |
| 1 | 1 | 0.0 | 1.0 | 1 | -1.0 | 0 | NaN | 1+ | 0 | 0.0 | 2H |
| 2 | 1 | 0.0 | 0.5 | 1 | 388800000.0 | 0 | NaN | 1/2+ | 1 | = 1.0000E+02 %B- ... | 3H |
| 3 | 1 | 0.0 | 0.5 | 1 | -1.0 | 0 | NaN | 1/2+ | 0 | 0.0 | 3He |
| 4 | 1 | 0.0 | 0.0 | 1 | -1.0 | 0 | NaN | 0+ | 0 | 0.000000 | 4He |
You can also load a limited set of levels based on the RIPL cut-off parameters by specifying the cutoff argument:
[15]:
ensdf_cutoff= nuc_data.load_ensdf(cutoff=True)
INFO:root:Reading data from C:/Users/Pedro/Desktop/ML_Nuclear_Data/ENSDF\CSV_Files/ensdf_cutoff.csv
[16]:
ensdf.shape
[16]:
(72667, 11)
[17]:
ensdf_cutoff.shape
[17]:
(21354, 11)
ML-ready RIPL Dataset¶
Modeling nuclear structure is tricky. There are many ways we could prepare the dataset and model it using different algorithms and approaches. One option is using processing similar to that of EXFOR. We can achieve this using the load_ensdf_ml() method.
NOTE: This method is experimental and the returned data should always be inspected.
[18]:
ensdf_ml, x_train, x_test, y_train, y_test, to_scale, scaler = nuc_data.load_ensdf_ml(
cutoff=False, log=True, append_ame=True, basic=1, num=True)
INFO:root:Dropping Ground State...
INFO:root:AME: Reading and loading Atomic Mass Evaluation files from:
C:/Users/Pedro/Desktop/ML_Nuclear_Data/AME/CSV_Files\AME_all_merged_no_NaN.csv
INFO:root:Dropping unnecessary features and one-hot encoding categorical columns...
INFO:root:Splitting dataset into training and testing...
INFO:root:Normalizing dataset...
INFO:root:Fitting new scaler.
INFO:root:Finished. Resulting dataset has shape (72138, 12), Training and Testing dataset shapes are (50496, 11) and (21642, 11) respesctively.
[19]:
x_train.head()
[19]:
| Level_Number | Z | N | A | Atomic_Mass_Micro | Mass_Excess | Binding_Energy | B_Decay_Energy | S(2n) | S(n) | S(p) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 2028 | 1.414973 | -1.618364 | -1.511965 | -1.553747 | -1.552763 | 1.144175 | -0.214506 | -3.507361 | 3.207512 | 3.279167 | 1.675159 |
| 63871 | 1.653213 | 1.208436 | 1.342290 | 1.292679 | 1.292933 | 0.767382 | -1.213664 | 1.179749 | -1.036423 | -1.285178 | 0.293675 |
| 45732 | 1.462398 | 0.480321 | 0.478158 | 0.479414 | 0.478960 | -0.595716 | -0.212229 | 0.989262 | -1.022829 | -0.572432 | -1.124873 |
| 58077 | 1.531479 | 0.951455 | 0.975689 | 0.967373 | 0.967281 | 0.121217 | -0.833843 | 0.635374 | -0.879278 | -1.059364 | -0.409278 |
| 58027 | 2.250420 | 0.951455 | 0.949503 | 0.951108 | 0.950984 | 0.065985 | -0.803468 | 0.114818 | -0.808407 | -0.424349 | -0.445178 |
This loader function offers less flexibility than the EXFOR counterpart. The same functionalities will hopefully be integrated in the future.