H-Matrix approximation for the operator exponential with applications
H-Matrix approximation for the operator exponential with applications
H-Matrix approximation for the operator exponential with applications
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
86 I. P. Gavrilyuk et al.<br />
η<br />
Γ 0<br />
Γ δ<br />
ξ<br />
δ 1<br />
γ 0<br />
γ 1<br />
Fig. 1. Parabolae Γ δ and Γ 0<br />
The boundedness of a(·, ·) implies <strong>the</strong> well-posedness of <strong>the</strong> continuous<br />
<strong>operator</strong> A : V → V ∗ defined by<br />
a(u, v) = V ∗< Au, v > V <strong>for</strong> all ∈ V.<br />
As usual, one can restrict A to a domain D(A) ⊂ V and consider A as an<br />
(unbounded) <strong>operator</strong> in H. The assumptions<br />
Re a(u, u) ≥ δ 0 ‖u‖ 2 V − δ 1‖u‖ 2 X<br />
|Im a(u, u)| ≤κ‖u‖ V ‖u‖ X<br />
<strong>for</strong> all u ∈ V,<br />
<strong>for</strong> all u ∈ V<br />
guarantee that <strong>the</strong> numerical range {a(u, u) :u ∈ X <strong>with</strong> ‖u‖ X =1} of<br />
A (and sp(A)) lies in Ω Γ0 , where <strong>the</strong> parabola Γ 0 depends on <strong>the</strong> constants<br />
δ 0 ,δ 1 ,κ,c e . Actually, if a(u, u) =ξ u + iη u <strong>the</strong>n we get<br />
ξ u = Re a(u, u) ≥ δ 0 ‖u‖ 2 V − δ 1 ≥ δ 0 c −2<br />
e − δ 1 ,<br />
|η u | = |Im a(u, u)| ≤κ‖u‖ V .<br />
It implies<br />
(2.2) ξ u >δ 0 c −2<br />
e − δ 1 , ‖u‖ 2 V ≤ 1 √<br />
ξ + δ1<br />
(ξ u + δ 1 ), |η u |≤κ .<br />
δ 0 δ 0<br />
The first and <strong>the</strong> last inequalities in (2.2) mean that <strong>the</strong> parabola Γ δ = {z =<br />
ξ + iη : ξ = δ 0<br />
κ<br />
η 2 − δ 1 } contains <strong>the</strong> numerical range of A. Supposing that<br />
Re sp(A) >γ 1 >γ 0 one can easily see that <strong>the</strong>re exists ano<strong>the</strong>r parabola<br />
Γ 0 = {z = ξ + iη : ξ = aη 2 + γ 0 } <strong>with</strong> a = (γ 1−γ 0 )δ 0<br />
(γ 1 +δ 1 )κ<br />
in <strong>the</strong> right-half<br />
plane containing sp(A), see Fig. 1. Note that δ 0 c −2<br />
e −δ 1 > 0 is <strong>the</strong> sufficient<br />
condition <strong>for</strong> Resp(A) > 0 and in this case one can choose γ 1 = δ 0 c −2<br />
e −δ 1 .<br />
Analogously to [4] it can be shown that inequality (2.1) holds true in C\Ω Γ0<br />
(see <strong>the</strong> discussion in [4, pp. 330-331]). In <strong>the</strong> following, <strong>the</strong> <strong>operator</strong> A is<br />
strongly P-positive.