From a615b93d9fad7a18d90604da7e192f39abd7a5c0 Mon Sep 17 00:00:00 2001 From: Denis Davydov Date: Mon, 4 Apr 2016 17:13:17 +0200 Subject: [PATCH] fix a typo in step-36 documentation --- examples/step-36/doc/intro.dox | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/examples/step-36/doc/intro.dox b/examples/step-36/doc/intro.dox index 799fd13e3a..9ebaa5954d 100644 --- a/examples/step-36/doc/intro.dox +++ b/examples/step-36/doc/intro.dox @@ -190,14 +190,14 @@ system that operates only on boundary nodes -- nodes that are not real degrees of freedom. Of course, since the two matrices $D_A,D_M$ are diagonal, we can exactly quantify these spurious eigenvalues: they are -$\varepsilon_{h,j}=A_{jj}/M_{jj}$ (where the indices +$\varepsilon_{h,j}=D_{A,jj}/D_{M,jj}$ (where the indices $j$ corresponds exactly to the degrees of freedom that are constrained by Dirichlet boundary values). So how does one deal with them? The fist part is to recognize when our eigenvalue solver finds one of them. To this end, the program computes and prints an interval within which these eigenvalues lie, by computing -the minimum and maximum of the expression $\varepsilon_{h,j}=A_{jj}/M_{jj}$ +the minimum and maximum of the expression $\varepsilon_{h,j}=D_{A,jj}/D_{M,jj}$ over all constrained degrees of freedom. In the program below, this already suffices: we find that this interval lies outside the set of smallest eigenvalues and corresponding eigenfunctions we are interested -- 2.39.5