Completeness of Gabor systems (Q281567): Difference between revisions

The authors study the question when a set of time-frequency shifts \[ \mathcal G(g,\alpha,\beta)=\left\{\text{e}^{2\pi \text{i} \beta j x}g(x-\alpha k)\,:\, j,k\in \mathbb Z\right \} \] is complete in \(L^2(\mathbb R)\), that is, the set of linear combinations of functions in \(\mathcal G(g,\alpha,\beta)\) is dense in \(L^2(\mathbb R)\). A related (but stronger) property that is discussed in the paper as well is that \(\mathcal G(g,\alpha,\beta)\) is a frame, i.e., \[ A\|f\|_2^2 \leq \sum_{j,k\in \mathbb Z} | \langle f, \text{e}^{2\pi\text{i} \beta j \cdot} g(\cdot-\alpha k)\rangle |^2 \leq B \|f\|_2^2, \] for some positive constants \(A\) and \(B\). The main result states that if \(g(x)= R(x)\text{e}^{-\gamma x^2}\) where \(\gamma >0\), \(R\) is either a rational function with no real poles or a finite sum of complex exponentials \(\sum_{m=1}^n c_m\text{e}^{\lambda_m x}\), and \(\alpha\beta\) is rational with \(\alpha \beta \leq 1\) then \(\mathcal G(g,\alpha,\beta)\) is complete. The authors also point out the vast difference between the completeness and the frame property for these Gabor system.