Beruflich Dokumente
Kultur Dokumente
ABSTRACT
In the present work, we have developed an intelligent mathematical model based on support vector machine
(SVM) for the investigation of the high spin band states for some odd mass nuclei (
, and
) which exhibit anomalies behavior in the moment of inertia. The comparison between our results and the
corresponding experimental ones manifests the same trend for the energy levels and backbending phenomenon and this
supports that our model could be applied successfully in the investigation, analysis and estimation of backbending for
odd mass nuclei.
KEYWORDS: Anomalies Behavior, Energy Levels, Estimation of Backbending
1. INTRODUCTION
The effect of back bending has been observed experimentally where the energetically favored, or yrast,
collective band under goes anabrupt increase in its moment of inertia (as a function of frequency, for example).
Original Article
Received: Oct 23, 2015; Accepted: Oct 28, 2015; Published: Nov 04, 2015; Paper Id.: IJPRDEC201501
The generally accepted interpretationis that back bending of an even-even nucleus occurs when two neutrons (or
protons) break their pairing bond and rotationally align perpendicular tothe symmetry axis. This back bending
has always attracted considerable experimental and theoretical attention especially for the even mass nuclei.
Furthe rmore, there are some nuclei exhibit a second anomaly in the moment of inertia. It was for the first time
measured for
158
Er[1],and the early interpretation was based on the alignment resulting from breaking of
aproton pair [2]. The back bending is also very intersting in case of the odd mass nuclei [3-8] and most of them
did not manupolate the calculations of back bending using an explicit mathematical equations. The description of
the backbending exprimental data of the odd mass nuclei still a problem of particular interest in the nuclear
structure physics. So, we are intersting to introduce a mathematical approachusing artificial intelligence (AI)
technique. Some odd mass nuclei with band structure at high spinstates exhibit a double back bending as
, second up bending as
9 12 .
In the present work with the advances in the field of artificial intelligence (AI) offers opportunities of
utilizing a new algorithms and models that enable researchers to solve the most complex systems [13-16]. In
recent years, support vector machine (SVM), which was introduced from statistical learning theory by Vapnik
[17], has received considerable attention and has been extensively used in many fields. Applying the SVM to
solve the regression problem is called support vector regression (SVR) [17, 18]. In this paper, the SVR is used as a
new approach to build the back bending model. The simulation results show that the SVR model has superior
www.tjprc.org
editor@tjprc.org
learning accuracy and generalization performance. Despite well-documented studies in other fields, the applications of
SVM in physics are few. D.O. Whiteson, et al. (2003) [13] applied SVM in the search for a theoretically predicted particle
and to the classi,cation of heavy quark jets. A. Vaiciulis et. Al. (2003) [ 14] used SVM for the case of identifying top quark
signal events in the dilepton decay channel amidst a large number of background events. S. Whiteson and D. Whiteson
(2006, 2009) [ 15, 16] presented a new approach that uses stochastic optimization techniques to directly search for
selectors that minimize statistical uncertainty in the top quark mass measurement
The objective of this study was to develop a new mathematical model using support vector regression
( SVR )-based modeling technique to studythe properties of the bands up tohighspins for the given nuclei.
Theyarelocatedonthe borderofthestronglydeformed region. . For determining the energy levels and the backbending up to
the second backbending of the given nuclei utilizing the SVR method. This study is the first to apply the SVR model
to calculate and predict the enegy levels and backbending of the odd mass nuclei.
The paper is organized as follows: In Section 2, the proposed modelling method of energy levels and backbending
is described. The results obtained are presented in Section 3. Finally, Section 4 provides the findings and conclusions.
2. THE PROPOSED SUPPORT VECTOR REGRESSION (SVR) MODEL FOR THE BACK-BENDING
SVR based algorithms use an implicit feature space mapping from the dimension of the data to a possibly infinite
feature space, providing a non-linear representation of the modeled data; this is done through the kernel trick (map every
data point into a higher dimensional space via some mapping function ( x ) ). One of the most common kernels with wide
applicability within SVR methods is the Radial Basis Function (RBF) kernel. As a result, it was also applied in this study.
When using the RBF kernel one has to choose the kernel width parameter, .
In this article, we use SVR with an insensitive loss function which penalizes the algorithm for deviations
larger than a threshold, . Another parameter, typically denoted C, sets the trade-off between the simplicity, and hence the
generalizability, and the allowed losses.
Considering a set of training dataset T = {(x1, y1) , ...., (xi, yi)}, where each xiRndenotes the input space of the
sample and has a corresponding target value yiR for i = 1, . . . , l where l corresponds to the size of the training data( in
our study, the input vector x={I, A, and Z} and the corresponding output is the energy level E(I, A, Z), where (I, A and Z)
are the spin , mass number and atomic mass of a nucleus, respectively. The idea of the regression problem is to determine a
function that can approximate future values accurately (Vapnik 1995)[17,18]. The generic SVR estimating function takes
the form in the higher dimensional space:
(1)
The SVR methodology reduces model error by incorporating the structural risk minimization (SRM) approach,
and also is seen as a form of regularization by a factor, C.
The factor, C, controls the trade-off between training error and model complexity (i.e., the size of model weights)
and must be chosen by the user.
This means that the support vector values ( SVR weights) are proportional to the error at each data point.
Given data, the user interested in developing an SVR model will have to specify the parameter pair (C, , ) and
,
then follow a principled routine to optimize the parameters thereafter.
With an appropriate modifying coefficient C, RBF band area width , and kernel function K, the optimum value
of each parameter can be solved by Lagrange. We follow Vapnik [17,18]and adopt the general form of the SVR-based
regression function defined as:
nSV
nSV
i=1
i=1
1
, (x p + x s ) , where x r and x s are anysupport vectors
2
i=1
that represent the input vector x = { I, A, and Z}
= ( i i * )x and b =
where
The final SVR function Eq (2), thus established is represented by the support vectors, name of the kernel used,
model parameters used (C, , ) non-zero Lagrange, (
of the weights w and bias term b. All these parameters are considered for calculating/predicting the new output(Energy
levels E(I,A,Z) based on any new input features in the final mathematical model (Eq. (2)).
Basically, it requires three steps to build an efficient SVR model: preparing training data, training the data to
obtain an SVR model, and predicting the new input data with the obtained model. The main purpose of creating and
using this model is to establish a model to predict the unseen experimental data.To determine the optimal parameters, we
employ the cross validation method, which is often used in practical applications. A k-fold cross-validation method was
applied to the experimental data sets for computing the validation of SVR model. In k-fold cross-validation method, the
data set is divided into k subsets, and the holdout method is repeated k times. At each time, k-1 subsets are used for
training and k-th subset is used for testing. Then the average error across all k trials is computed. Therefore, every data
point gets to be in a test set exactly once, and gets to be in a training set k-1 times.
Some statistical metrics such as, normalized mean square error (NMSE), mean square error (MSE), and the
correlation coefficient (CC) were used to evaluate the prediction performance of the two models.NMSEand the correlation
coefficient were used to measure the deviation between the actual and calculated values. The smaller the values of NMSE
and MSE, the closer were the predicted values to the actual values. A higher CC number means a better model, with a (1)
meaning perfect statically correlation and a (0) meaning there is no correlation at all.
The energies E(I,A,Z) of the yrast band as a function of spin (I), mass number (A) and atomic number (Z) for the
given nuclei can be calculated from equation (2). Also for a better understanding of the multiple backbending phenomena,
the theoretical results and the experimental data are compared by means of backbending plots and the corresponding
energy spectra.The backbending plotisagraph which shows the dependence of the backbending plots on calculations of
the moment of inertia #(%) according to the following expression:
www.tjprc.org
editor@tjprc.org
*'+
(3)
,('- ,/,0)+,(',/,0)
2,(',/,0)
2'
,(' + , /, 0) ,(', /, 0)
(4)
&(')
, there
appears a gradual increase in the moment of inertia in the lower energy states, then a radical change in behavior and then
return again to the gradual increase once more. This effect is known as a first backbending which is ocucure in some heavy
nuclei due to the breaking of couple pair of nucleon. When this effect occurs, the unpaired nucleons go into different orbits
and change the moment of inertia. Sometimes this happens again at heigher energy states to give the second backbendingin
nuclei.
In this work, we developed an intelligent mathematical model to calculate and predicthe energy levels of the odd
mass nuclei. The standard k-fold cross-validation (10-fold cross-validation) was carried out for the estimation of the best
model parameters (C, , ) . In this approach, for a given combination of width of RBF kernel ( ), cost coefficient (C)
and loss function parameter ( ), the training set is first randomly divided into k (10) equal sized subsets. Next, k number
of models are constructed by leaving out a different subset each time, with the remaining (k-1=9)subsets collectively
representing the training set.
An average of the error correspondingto the left-out subsets, known as cross-validation error gives an estimate
of the model performance if a large-sized data set was available for building the model. After evaluation of the model for
the wide range of model parameters (grid search methodology), gives the optimal values of them. So obtained optimal
= 10.576826
values of the three model parameters, corresponding to BB model ((Parameter values: C = 327.494565
= 0.000480)). Optimally selecting these model parameters, automatically decides over the optimal number of support
vectors (65), which plays a vital role in the performance of the SVR -based model. The developed SVR
modeldemonstrates a good performance in predicting the energy levels and backbending , given the minimum root mean
square error (NMSE = 0.00102) and maximum correlation coefficient (CC = 0.995) during the prediction periods.
For the application of this model to calculate the energy levels and backbending we studied
and
,
+
. The energy levels calculated from the present work compared to the experimental
data are reproduced with high accuracy as shown in figure 1(a,b,c,d,e) and in table 1. As can be seen in figure 2(a) The plot
of I(I+1) vs E(I) for the experimental data of
and
positive parity states which denotes that we have two band crosssing approximatly at 11~0.24 < = and
12~0.3 < =. In figure 2.(b) there are also two deviations in the path of the curve for
and
approximatly at spins ~
positive parity states which denotes that there are two band crossing in this nucleus approximatly at11~0.36 < =
and
and
for each one of the two nuclei, respectively, as shown in figure 2(c,d). These results denotes that
we have two band crossing for them approximatly at11~0.34 < =, 12~0.48 < = for
12~0.49 < = for
. The results of these four nuclei for the two band crossing means that we have double
backbending in the experimental data which is shown in figure 3(a,b,c,d). Furthermore, the last nucleus
, it has only
positive parity state approximatly at11~0.27 MeV which denotes that there is one band
on deviation at spin~
crossing i.e There is only one backbending in calculation which are shown in figure 3(e). The backbending calculations
using our predicted equation gives an excellent agreement results compared to the experimental values as shown in figure
3(a,b,c,d,e).
O riginP ro 8 Evaluation
O riginPro 8 E valuation
O riginP ro 8 Evaluation
O riginPro 8 E valuation
O riginP ro
8 E valuation
O riginP ro
8 E valuation
O riginP ro
8 E valuation
O riginPr o 8 E valuation
O riginP ro
8 E valuation
O riginPr o 8 E valuation
O riginP ro
8 E valuation
O riginPr o 8 E valuation
O riginP ro
8 E valuation
O riginPr o 8 E valuation
10
10
O riginPr o 8 E valuation
171
75
Re
10
O riginP ro 8 Evaluation
163
O riginPro 8 E valuation
72
8
O riginP ro 8 Evaluation
O riginPro 8 E valuation
E(I)
E(I) (MeV)
(MeV)
12
O riginP ro 8 Evaluation
O riginPro 8 E valuation
O riginP ro 8 Evaluation
O riginPro 8 E valuation
4
4
EXPERIMENT
THEORY
10
15
20
25
30
10
15
20
(h)
25
(a)
10
O riginP ro
8 Evaluation
O riginP ro
8 Evaluation
165
72
Hf
O riginP ro
8 Evaluation
E(I)
E(I)
10
8 Evaluation
6
Orig inPro 8 Eva luation
35
12
161
O riginP ro
30
40
(h)
(b)
Experiment
Theory
O riginPr o 8 E valuation
Hf
72
Hf
4
Orig inPro 8 Eva luation
O riginP ro
8 Evaluation
EXPERIMENT
THEORY
2
Orig inPro 8 Eva luation
O riginP ro
8 Evaluation
EXPERIMENT
THEORY
10
15
20
25
30
35
10
15
(h)
20
25
(c)
30
35
40
(h)
(d)
6
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
O rig in P ro
E v a l u a tio n
O ri g in P ro
E v a l u a tio n
E(I)
MeV)
181
75
Re
0
4
10
12
14
16
EXPERIMENT
THEORY
18
20
22
24
(h )
()
Figure 1: (a,b,c,d,e) Demonstrates the Relation between the Spin and the
Calculated Energy (Theory) using Our Predicted Equation Compared to the Experiment
Energy Data (Experiment) [9-12] for the Nuclei 171Re,163Hf, 165Hf, 161Hf, and 181Re
www.tjprc.org
editor@tjprc.org
O rigi nP ro 8 Evaluati on
12
OriginPro 8 Evaluation
6
171
10
O rigi nP ro 8 Evaluati on
3
2
O rigi nP ro 8 Evaluati on
O rigi nP ro 8 Evaluati on
O rigi nP ro 8 Evaluati on
163
OriginPro 8 Evaluation
O riginPro 8 Evaluation
Hf
72
(MeV)
E(I)
E(I)
(MeV)
O riginPro 8 Evaluation
O rigi nP ro 8 Evaluati on
Re
75
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
EXPERIMENT
EXPERIMENT
200
10
800
1000
200
400
600
800
1000
I(I+1)
(a)
(b)
O riginPro 8 Evaluation
O riginPro 8 Evaluation
165
72
Hf
12
OriginPro 8 Evaluation
10
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
OriginPro 8 Evaluation
72
O riginPro 8 Evaluation
O riginPro 8 Evaluation
(MeV)
1400
1600
OriginPro 8 Evaluation
161
O riginPro 8 Evaluation
1200
OriginPro 8 Evaluation
Hf
E(I)
(MeV)
600
I(I+1)
E(I)
400
2
O ri gi nPro 8 Eval uation
O riginPro 8 Evaluation
EXPERIMENT
EXperimental
0
200
400
600
800
1000
1200
1400
200
400
600
800
I(I+1)
I(I+1)
(c)
(d)
1000
1200
1400
6
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
181
(MeV)
E(I)
75
Re
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
O riginPro 8 Evaluation
OriginPro 8 Evaluation
Experiment
0
0
100
200
300
400
500
600
I(I+1)
(e)
Figure 2: (a,b,c,d,e) Demonstrates the Relation between the I(I+1) and Experimental Energy Data
(Experiment) for the Nuclei 171Re,163Hf, 165Hf, 161Hf, and 181Re. It Deviate at One Or
Two Points which Denots One Band Crossing Or Two Band Crossing for the Specified Nuclei
160
O r i g i n Pr o 8 E v a l u a t i o n
O r i g i n Pr o 8 Ev a l u a t i o n
180
Or ig in Pro 8 Ev a lu at io n
120
Or ig in Pro 8 Ev a lu at io n
Or ig in Pro 8 Ev a lu at io n
Or ig in Pro 8 Ev a lu at io n
Or ig in Pro 8 Ev a lu at io n
Re
163
Hf
O r i g i n Pr o 8 E v a l u a t i o n
140
100
80
(MeV)-1
(MeV)-1
(2/h2)
Or ig in Pro 8 Ev a lu at io n
75
140
150
8 Eval uati on
130
(2/h2)
160
Ori gi nPro
171
110
72
O r i g i n Pr o 8 Ev a l u a t i o n
O r i g i n Pr o 8 E v a l u a t i o n
O r i g i n Pr o 8 Ev a l u a t i o n
O r i g i n Pr o 8 E v a l u a t i o n
O r i g i n Pr o 8 Ev a l u a t i o n
O r i g i n Pr o 8 E v a l u a t i o n
O r i g i n Pr o 8 Ev a l u a t i o n
O r i g i n Pr o 8 E v a l u a t i o n
O r i g i n Pr o 8 Ev a l u a t i o n
120
100
60
40
20
0.00
0.02
0.04
0.06
0.08
2
(h)
0.10
0.12
2
(MeV)
(a)
90
EXPERIMENT
THEORY
0.14
0.16
EXPERIMENT
THEORY
80
0.18
0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26
2
(h) (MeV)
(b)
140
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
O r i gi nP r o 8 Ev al uat i o n
Or i g i nPr o 8 Eva l ua t i on
(2/h2)
(MeV)-1
130
120
165
72
Hf
110
EXPERIMENT
THEORY
100
90
0.10
0.15
0.20
2
(h)
0.25
0.30
(MeV)
(c)
Figure 3: (a,b,c,d,e) Demonstrates Thebackbending Plots for the First and the
Second Backbending which Arewell Reproduced in Our Calculations (THEORY) with the
Experimental Data (EXPERIMENT) [9-12] for the Inset Nuclei
Table 1: Comparison between our Results (Etheo) and Different Experimental (Eexp) [9-12]
(for the Band Energies of 171Re, 163Hf, 165Hf, 161H1, and 181Re isotopes).
All the Energy Values are in MeV
171
Nucleus
5/2+
9/2+
13/2+
17/2+
21/2+
25/2+
29/2+
33/2+
37/2+
41/2+
45/2+
49/2+
53/2+
57/2+
61/2+
www.tjprc.org
Re
Eexp Etheo
-0.365 0.365
0.773 0.772
1.206 1.256
1.678 1.677
2.042 2.03
2.406 2.412
2.911 2.911
3.514 3.514
4.156 4.133
4.713 4.713
5.29 5.298
5.977 5.977
6.766 6.767
--
163
Hf
Eexp Etheo
----0.841 0.841
1.405 1.404
2.063 2.066
2.793 2.793
3.556 3.54
4.271 4.266
4.964 4.973
5.705 5.705
6.511 6.511
7.385 7.387
8.222 8.288
165
Hf
Eexp Etheo
-----1.224 1.224
1.825 1.825
2.485 2.486
3.167 3.166
3.843 3.845
4.532 4.533
5.272 5.267
6.079 6.078
6.961 6.968
7.913 7.913
161
Hf
Eexp Etheo
----1.138 1.138
1.698 1.699
2.324 2.315
3.002 3.002
3.73 3.751
4.519 4.519
5.274 5.269
6.007 6.011
6.79 6.79
7.639 7.638
8.553 8.558
181
Re
Eexp Etheo
-0.267 0.267
0.646 0.646
1.116 1.120
1.642 1.642
2.177 2.175
2.71 2.711
3.272 3.273
3.904 3.90
4.612 4.613
5.386 5.386
-----
editor@tjprc.org
65/2+
69/2+
73/2+
77/2+
-----
Table 1: Contd.,
9.172 9.171 8.882 8.895
10.051 10.053 9.928 9.927
10.986 10.987
-11.972 11.972
--
9.531 9.531
10.546 10.546
11.599 11.599
-----
In the study, we investigated the ability of the support vector regression (SVR), which is mathematically wellfounded and provides new insights into function approximation. SVR demonstrates a good performance in predicting BB,
given the minimum root mean square error (NMSE = 0.00102) and maximum correlation coefficient (CC = 0.995) during
the prediction periods.
4. CONCLUSIONS
This paper propose an SVR based computational model to calculate and predict the energy levels and
backbending of a deformed odd mass nuclei171Re, 163Hf, 161Hf, 165Hf , and 181Re . These nuclei exhibit a double anomaly in
the moment of inertiafor the first four nuclei and only one anomaly for the last nucleus. This anomaly arises from the
broken pair of neutron to give the first band crossing at a certain value (11) and a second band crossing caused by a
successive breaking of proton pair at another approximatly value (12). The results of our model SVR has high
prediction accuracy and a excellent agreement with the energy states as well as the backbending behavior of the odd mass
nuclei.
Application of the proposed models in the prediction of backbending for odd mass nuclei will allow to obtain a
satisfactorily accurate understanding of the backbendingin addition to the confirmation of the experimental results.This
proves its utility as a design estimation tool for nuclear physics.
REFRENCES
1.
2.
3.
S. MohammediandR. Shaheldare, Asian Journal of Engineering and Technology Innovation 02(02),(01-04) (2014).
4.
R. Bengtsson, http://hal.archives-ouvertes.fr/jpa-00220628.
5.
6.
7.
8.
9.
https://www.nds.iaea.org/ - ENSDF evaluation: Coral M. Baglin Nucl. data sheets 96,399 (2002).
10. [https://www.nds.iaea.org/ - ENSDF evaluation: Balraj Singh atomic and nuclear data tables 99,69(2013).
11. https://www.nds.iaea.org/ - ENSDF evaluation: Ashok K. Jain, Anwesha Ghosh and Balraj Singh Nucl. data sheets
107,1075(2006).
12. https://www.nds.iaea.org/ - ENSDF evaluation: S.-C.Wu Nucl. data sheets 106,367 (2005).
13. D.O. Whiteson, N.A. Nauman; Support vector regression as a signal discriminator in high energy physics; Neurocomputing
www.tjprc.org
editor@tjprc.org