We all know that a Gram matrix (a matrix with entries that are inner products of basis functions) is a invertible. Suppose I have $a_{ij} = (h_i, h_j)_H$ where the $h_j$ are basis functions of a Hilbert space $H$. Is the infinite matrix $A=(a_{ij})$ invertible? That is, is $\sum_{j=1}^\infty a_{ij}v_i = b_i$ solvable for $v_i$ for each $i$ given $b_i$?



For an arbitrary element $x$, let $x^* \in H^*$ denote $(\cdot,x)_H$.

Let $V$ denote the transformation $$ Vx = \sum_{i} x_i h_i $$ That is, $V$ is the matrix whose "column vectors" are the basis functions $h_i$. Note that $A = V^*V$.

Now, note that for any $x$, $x^*(Ax) = (x^* V^*)(Vx) = (Vx)^*(Vx) = \|Vx\|_H^2$. It follows that $Ax = 0$ if and only if $Vx = 0$. Since $V$ has a basis as its columns, $V$ is indeed invertible.

  • $\begingroup$ thanks but I am confused about what $V$ is. $V:H \to H$ but how does $x_ih_I$ make sense?? $\endgroup$
    – user173805
    Sep 8 '14 at 21:03
  • $\begingroup$ Select an orthonormal basis $\{v_i\}$. Define $V : H \to H$ by $\sum x_i v_i \mapsto \sum x_i h_i$; that is, by $v_i \mapsto h_i$. $\endgroup$ Sep 8 '14 at 21:07
  • $\begingroup$ ...........thank u $\endgroup$
    – user173805
    Sep 8 '14 at 21:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.