A geometric spectral theory for \(n\)-tuples of self-adjoint operators in finite von Neumann algebras (Q1296770)

Consider a finite von Neumann algebra \(M\) with a faithful trace \(\tau\), and selfadjoint \(b_1,\dots, b_n\in M\). The spectral scale \(B\) of \(\tau,b_1,\dots, b_n\) is defined as the image of the positive unit ball of \(M\) under the map \(\Psi: M\to\mathbb{C}^{n+1}\), \(\Psi(x)= (\tau(x),\tau(xb_1),\dots, \tau(xb_n))\). The paper uncovers the relevance of the geometry of the convex set \(B\subset \mathbb{R}^{n+1}\) for the spectral properties of the \(n\)-tuple \(b_1,\dots, b_n\). In the case \(n=1\), the slopes of the lower boundary of \(B\) are the spectrum of \(b_1\) and the spectral projections of \(b_1\) can be recovered as preimages of \(\Psi\) (Theorems 1.5-7). For general \(n\), \(B\) depends only on the restriction of \(\Psi\) to \(N= \{b_1,\dots, b_n\}''\) (Theorem 2.11) and provides the spectral information for all linear combinations of \(b_1,\dots, b_n\) (Theorem 2.3). If \(N\) is Abelian, \(N\) determines \((\tau,N)\), meaning that any two \((\tau,N)\)'s with the same \(B\) have equivalent tracial representations (Theorem 3.4). This is no longer true if \(N\) is non-Abelian, but a matricial version of \(B\), the complete spectral scale, is shown to take over the property (Theorem 3.11).