pdf, 9 MiB - Infoscience - EPFL
pdf, 9 MiB - Infoscience - EPFL
pdf, 9 MiB - Infoscience - EPFL
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
42 CHAPTER 2. NUMERICAL METHODS<br />
The following change in the eigenvalues are obtained with first order perturbation<br />
theory:<br />
λ i = λ 0i + x T 0i ([δK]) x 0i, (2.53)<br />
and the eigenvectors are given by:<br />
x i = x 0i + ∑ j≠i<br />
x T 0i ([δK]) x 0j<br />
λ 0i − λ 0j<br />
x 0j (2.54)<br />
The derivatives are easily obtained from these formulae. In the case of degenerate<br />
eigenvectors, the above formulae are no longer valid, and a more general<br />
degenerate perturbation theory should be considered.<br />
Research on the derivative of degenerate eigenvectors has been an area of significant<br />
interest in recent years. Engineer have focused on the study of the so<br />
called sensitivities, that are simply the derivatives of eigenvalues and eigenvectors<br />
with respect to design parameters that make the matrix evolve. Degenerate eigenvalues<br />
are unavoidable in practice, and, for instance, appear in the calculation of<br />
the derivatives of the eigenvectors of the BCS Hamiltonian.<br />
We sketch only here the main results obtained for the derivate of a degenerate<br />
eigenvector (for more details the reader is referred to Ref. [53,54,55,56]). Let us<br />
assume without loss of generality that the first eigenvalue λ 1 is degenerate with<br />
multiplicity r. We define Φ as the matrix containing all the eigenvectors, Φ 1 as<br />
the matrix of the degenerate eigenvectors, and Φ 2 as the matrix containing the<br />
other vectors. With this definitions we have the trivial identity:<br />
[ ]<br />
Φ T λ1 1<br />
K 0 (p)Φ=Λ= r 0<br />
, (2.55)<br />
0 Λ 2<br />
It can be proved that the derivative of the degenerate eigenvectors φ 1 is :<br />
Φ ′ 1 =Φ 2 (λ 1 I − Λ 2 ) −1 Φ T 2 D jΦ 1 (2.56)<br />
Knowing the derivative of the U and V matrices allows to measure the O k observables<br />
of equation (2.40) and the stochastic minimization procedure is now<br />
well defined.<br />
2.3.3 Correlated measurement minimization<br />
One limitation of the stochastic minimization method, in our present implementation,<br />
is that the formulae obtained for the derivation of a determinant are not<br />
valid for the more general Pfaffian wavefunction, since there is no simple formula<br />
for the derivative of a Pfaffian (see equation (2.48)). Therefore, in the case<br />
of a Pfaffian Monte Carlo simulations, the gradients of the energy with respect<br />
to the variational parameters have to computed numerically. This task is very<br />
heavy and limits the minimization to a few parameters (about 10 parameters at