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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

where ē is an arbitratry vector. Denoting a given function of matrix elements by S( ¯M) we have to locate thepoints ¯M on the contour yielding the extremum values of S( ¯M). Let us consider the function S( ¯M) whichcan locally be appoximated by the linear expansion in ¯M:∆ s ( ¯M) =∇ s • ( ¯M − ¯M s ) (6.33)where we exp<strong>and</strong> S( ¯M) in a vicinity of ¯M 2 .Tofind the extrema of 6.33 on the coutour 6.30 we have tofind the vectors ē satisfying:µd 2δ • ( ¯∇ s • ē)=0 (6.34)dē ¯∇ o ē +[(¯∇ o ē) 2 +2δēJē] 1/2which yields:¯∇ s (2δ − ¯∇ o ē) − ¯∇ o ( ¯∇ s ē) − ( ¯∇ s ē) • ¯Je =0 (6.35)From 6.35 it is clear that the vector ē must be a linear combination of J −1 ¯∇s <strong>and</strong> J −1 ¯∇o .Insertingto 6.35 we get, using the identityē = αJ −1 ¯∇s + βJ −1 ¯∇o (6.36)¯∇ s J −1 ¯∇o = ¯∇ o J −1 ¯∇s (6.37)resulting from the symmetry of J :µ 2δ + ¯∇o J −1 1/2¯∇0α = ±¯∇ s J −1 ¯∇ ; β = −1 (6.38)swhich is valid for any arbitrarily chosen origin ¯M o . However, in our case we can assume that ¯M o ,thevector of fitted matrix elements, is a close approximation of the minimum, thus ¯∇ o ≈ 0. Neglecting theterms containing ¯∇ o we finally get:µ 1/2¯M = ¯M 2δo ±¯∇ s J −1 ¯∇ J −1 ¯∇s (6.39)swhere a positive sign corresponds to the maximum of 6.33 on the contour 6.30 <strong>and</strong> a negative signcorresponds to its minimum. This formula gives an exact solution for any linear function of ¯Mo . Fornon-linear functions 6.39 can be used iteratively following the scheme:¯∇ (0)s = ¯∇ s ( ¯M o )=⇒¯M(1)¯∇ (1)s = ¯∇ s ( ¯M (1) )=⇒ ¯M (2) (6.40)until the convergence is achieved. This procedure is used in SIGMA to estimate the errors of Q 2 , cos3δ<strong>and</strong> their statistical moments. Because of the fact that the implementation of the rotational-invariant sumrules is only possible for the cases in which virtually all the E2 matrix elements for low lying states areknown, implying that the underlying Coulomb <strong>excitation</strong> problem is well overdetermined, it is reasonable toassume a least-squares statistic increase δ= 1.TheJ matrix is estimated using the gradients computed byGOSIA during the calculation of correlated errors (see 4.6) <strong>and</strong> stored on a permanent file. Applying thequadratic approximation 6.1 <strong>and</strong> assuming ¯M o = minimum (i.e. ∇ o =0)onecanwrite:¯∇( ¯M) =Ĵ( ¯M − ¯M o ) (6.41)During the correlated errors calculation GOSIA evaluates the gradients in points -M for which only onematrix element at the time is perturbed from its central value. This means that ¯M − ¯Mo has only onenon-zero component, thus 5.41 defines k-thcolumnofJ if M k has been perturbed. Two estimates of J ki125

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

Saved successfully!

Ooh no, something went wrong!