Notes about math, research, and more.
:tags: learning-k-graphs

Chapter 1

  • Cuntz-Krieger family and the Cuntz-Krieger relations
  • Since projections \(P_{v}\) are mutually orthogonal, the ranges \(P_{v}\mathcal{H}\) are mutually orthogonal subspaces of \(\mathcal{H}\).
  • The first CK relation says \(S_{e}\) has initial space \(P_{s(e)}\mathcal{H}\)
  • Second relation implies that the range projection \(S_{e}S^*_{e}\) is dominated by \(P_{r(e)}\), hence \(S_{e}\mathcal{H} \subset P_{r(e)}\mathcal{H}\).
  • \(S_{e}\) is an isometry of \(P_{s(e)}\mathcal{H}\) onto a closed subspace of \(P_{r(e)}\mathcal{H}\). This gives

\[ S_{e} = P_{r(e)}S_{e} = S_{e}P_{s(e)} \] which is very important function.

  • Further, CK2 implies that \(S_{e}\) for each edge \(e\) with \(r(e) = v\) have mutually orthogonal ranges with span \(P_{v}\mathcal{H}\) so

\[ P_{v}\mathcal{H} = \bigoplus_{\{ e \in E^1 : r(e) = v \}} S_{e}\mathcal{H} \]

Finding Cuntz-Krieger $E$-families with \(P_{v}, S_{e}\) nonzero

Take \(H_{v}\) to be a separable infinite-dimensional Hilbert space for each \(v \in E^0\), set \(\mathcal{H} = \bigoplus_{v} \mathcal{H}_{v}\). Then take \(P_{v}\) to be the projection of \(\mathcal{H}\) on \(\mathcal{H}_{v}\), decompose \(\mathcal{H}_{v}\) as a direct sum \(\mathcal{H}_{v} = \bigoplus_{r(e) = v} H_{v,e}\)

Examples

\begin{tikzpicture}[->,>=stealth',shorten >=1pt,auto,node distance=2cm, thick,main node/.style={circle,draw,font=\sffamily\Large\bfseries}] \node[main node] (1) {v}; \node[main node] (2) [right of=1] {w}; \path[every node/.style={font=\sffamily\small}] (1) edge [loop above] node {$e$} (1) (2) edge node {$f$} (1); \end{tikzpicture}