12.07.2015 Views

coulomb excitation data analysis codes; gosia 2007 - Physics and ...

coulomb excitation data analysis codes; gosia 2007 - Physics and ...

coulomb excitation data analysis codes; gosia 2007 - Physics and ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

4.5 MinimizationThe minimization, i.e. fitting the matrix elements to the experimental <strong>data</strong> by finding a minimum of theleast-squares statistic is the most time-consuming stage of Coulomb <strong>excitation</strong> <strong>analysis</strong>. A simultaneous fitof a large number of unknown parameters (matrix elements) having in general very different influences onthe <strong>data</strong>, is a complex task, which, to whatever extent algorithmized, still can be slowed down or speeded updependent on the way it is performed. GOSIA allows much freedom for the user to define a preferred strategyof the minimization. A proper use of the steering parameters of the minimization procedure can significantlyimprove the efficiency of fitting dependent on the case analyzed - in short, it still takes a physicist to get theresults!. The following overview of the fitting methods used by GOSIA is intended to provide some ideasabout how to use them in the most efficient way.4.5.1 Definition of the Least-square StatisticA set of the matrix elements best reproducing the experimental <strong>data</strong> is found by requesting the minimumof the least-squares statistic, S( ¯M). The matrix elements will be treated as a vector ordered according tothe user-defined sequence. The statistic S is in fact a usual χ 2 -type function, except for the normalizationto the number of <strong>data</strong> points rather than the number of degrees of freedom which cannot be defined due toaverydifferent sensitivity of the <strong>excitation</strong>/de<strong>excitation</strong> process to various matrix elements, as discussed inmore detail in Section 4.6. The statistic S, called CHISQ in the output from GOSIA, is explicitly given by:ÃS(M) = 1 S y + S 1 + X !w i S i (4.24)Niwhere N is the total number of <strong>data</strong> points ( including experimental yields, branching ratios, lifetimes,mixing ratios <strong>and</strong> known E2 matrix elements ) <strong>and</strong> S y , S 1 <strong>and</strong> S i are the components resulting from varioussubsets of the <strong>data</strong>, as defined below. Symbol w i st<strong>and</strong>s for the weight ascribed to a given subset of <strong>data</strong>.The contribution S y to the total S function from the measured γ-yields following the Coulomb <strong>excitation</strong>,is defined as:S y = X I e I dw IeI dX1σ 2 k(I e,I d ) k(C IeI dY ck − Y ek ) 2 (4.25)where the summations extend over all experiments (I e ), γ-detectors (I d ) <strong>and</strong> experiment- as well as detectordependentobserved γ-yields, indexed by k. The weights ascribed to the various subsets of <strong>data</strong> (w IeI d)canbe chosen independently for each experiment <strong>and</strong> γ-detector, facilitating the h<strong>and</strong>ling of <strong>data</strong> during theminimization (particularly, some subsets can be excluded by using a zero weight without modifying theinput <strong>data</strong>). The superscript “c“ denotes calculated yields, while the superscript “e“ st<strong>and</strong>s for experimental<strong>data</strong>, with σ k being the experimental errors. The coefficients C Ie I dare the normalization factors, connectingcalculated <strong>and</strong> experimental yields (see 4.5.2).The next term S 1 of equation 4.24 is an “observation limit“ term, intended to prevent the minimizationprocedure from finding physically unreasonable solutions producing γ-ray transitions which should be, buthave not been, observed. An experiment <strong>and</strong> detector dependent observation upper limit, u(I e,I d ) is introduced,which is the ratio of the lowest observable intensity to that of the user-specified normalizationtransition (see V.29). S 1 is defined as:S 1 = X jµ Yc 2j (I e ,I d )1Yn c − u(I e ,I) ·(I e ,I d) u 2 (I e ,I d )(4.26)the summation extending over the calculated γ transitions not defined as experimentally observed <strong>and</strong>covering the whole set of the experiments <strong>and</strong> γ-detectors defined, provided that the upper limit has beenexceeded for a given transition. The terms added to S 1 are not counted as <strong>data</strong> points, thus the normalizationfactor N is not increased to avoid the situation in which the fit could be improved by creating unreasonabletransition intensities at the expense of the normalization factor.The remaining terms of 4.24 account for the spectroscopic <strong>data</strong> available, namely branching ratios, meanlifetimes, E2/M1 mixing ratios <strong>and</strong> known E2 matrix elements. Each S i term can be written as:43

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!