1
$\begingroup$

I'm reading an article regarding the condition for the existence of a solution to a constrained Pick interpolation problem, and the author wrote the following equality - for background, we have the following spaces: $$H^2_{\alpha,\beta}\subset M_k(H^2)$$ where $M_k(H^2)$ is the set of $k\times k$ matrices with entries being functions in $H^2$, and inner product given by: $$\langle f,g\rangle=\frac{1}{2\pi}\int_{\mathbb{T}}tr[f(z)g(z)^*]|dz|$$

Now given points $z_1,\ldots,z_n$, we let: $$B_FM_k(H^2)=\{f\in M_k(H^2):f(z_i)=0,\forall i=1,\ldots,n\}$$

To the question: the author defines the subspace: $$M_{\alpha,\beta}=H^2_{\alpha,\beta}\cap(H^2_{\alpha,\beta}\cap B_FM_k(H^2))^{\perp}\subset H^2_{\alpha,\beta}$$ Now assuming that $Q\in M_k(H^{\infty})$ is such that $M_Q$ (the multiplier operator by $Q$) is invariant for $H^2_{\alpha,\beta}$, $g\in H^2_{\alpha,\beta}$ and $f\in (H^2_{\alpha,\beta}\cap B_FM_k(H^2))^{\perp}$, the author claims that: $$\langle Qg,f\rangle=\langle P_{H^2_{\alpha,\beta}}Qg,P_{(H^2_{\alpha,\beta}\cap B_FM_k(H^2))^{\perp}}f\rangle=\langle P_{M_{\alpha,\beta}}Qg,P_{M_{\alpha,\beta}}f\rangle$$

I understand the first equality, but cannot understand the second equality. Any ideas or explanations on why this is true would be appreciated. (Clarification: $P_W$ is the orthogonal projection on the subspace $W$).

$\endgroup$

1 Answer 1

1
$\begingroup$

I don't fully understand the context here, but it seems to me what you need is the following: Let $A = H^2_{\alpha, \beta}$, $B = (H^2_{\alpha, \beta} \cap B_FM_k(H^2))^\perp$, then $P_BP_A = P_{A \cap B}$. We can show that this would happen if $A \cap (A \cap B)^\perp \perp B$. (Indeed, we observe that $P_BP_A$ and $P_{A \cap B}$ both reduce to 1 on $A \cap B$, so it suffices to show $P_BP_A((A \cap B)^\perp) = 0$. Fix any $h \in (A \cap B)^\perp$, write $h = P_Ah + P_{A^\perp}h$. Since $A^\perp \subseteq (A \cap B)^\perp$, we have, $P_Ah = h - P_{A^\perp}h \in (A \cap B)^\perp$, so $P_Ah \in A \cap (A \cap B)^\perp$. As $A \cap (A \cap B)^\perp \perp B$, we have $P_BP_Ah = 0$.) Note that in this case, $(A \cap B)^\perp = \overline{A^\perp + B^\perp} = \overline{(H^2_{\alpha, \beta})^\perp + (H^2_{\alpha, \beta} \cap B_FM_k(H^2))} = (H^2_{\alpha, \beta})^\perp + (H^2_{\alpha, \beta} \cap B_FM_k(H^2))$, as the two summands are orthogonal to each other, so $A \cap (A \cap B)^\perp = H^2_{\alpha, \beta} \cap ((H^2_{\alpha, \beta})^\perp + (H^2_{\alpha, \beta} \cap B_FM_k(H^2))) = H^2_{\alpha, \beta} \cap B_FM_k(H^2) = B^\perp$. This proved the claim.

$\endgroup$
3
  • $\begingroup$ Is there a general theorem for when $P_BP_A=P_{A\cap B}$? $\endgroup$
    – Math101
    Commented Aug 28, 2023 at 9:07
  • $\begingroup$ @GBA The condition I wrote down is an equivalent condition, as should be relatively easy to verify. By taking adjoints you can see that it’s also equivalent to $P_AP_B = P_{A \cap B}$ and therefore also equivalent to $B \cap (A \cap B)^\perp \perp A$. If you’re asking for equivalent conditions, that’s all I’m aware of. $\endgroup$
    – David Gao
    Commented Aug 28, 2023 at 16:01
  • $\begingroup$ A generalization of the argument I used will show that $B^\perp \subseteq A$ (or equivalently $A^\perp \subseteq B$) is a sufficient condition, which applies to your question, though this is not a necessary condition. $\endgroup$
    – David Gao
    Commented Aug 28, 2023 at 16:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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