If $\{\boldsymbol{\varphi}_1, \boldsymbol{\varphi}_2, ... \}$ is an orthonormal system in $\mathcal{H}$, then the Bessel's inequality is:

$$\sum_{j=1}^{\infty} |\langle \mathbf{x},\boldsymbol{\varphi}_j \rangle|^2 \leq \lVert \mathbf{x} \rVert^2 \quad \text{for every } \mathbf{x}\in \mathcal{H}.$$

My question is:

  • What is the meaning/concept of this inequality?
  • Is there any geometric or physical interpretations about that?

Any comments or answers are appreciated.

  • 4
    $\begingroup$ The idea is that, if you had a full basis, you would expect $x=\sum_j \langle x,\varphi_j\rangle\varphi_j$, which would then lead to $\|x\|^2=\sum_j |\langle x,\varphi_j\rangle|^2$. If the system is orthonormal, but not necessarily a full basis, then $\sum_j |\langle x,\varphi_j\rangle|^2 \le \|x\|^2$, because you might be missing components in the sum $\sum_j \langle x,\varphi_j\rangle \varphi_j$ needed to have a full expansion of $x$. $\endgroup$ – DisintegratingByParts Apr 17 '17 at 10:56
  • $\begingroup$ Thanks @TrialAndError. Could you please explain about the orthonormal basis and full basis in $\mathcal{H}$? $\endgroup$ – Amin Apr 17 '17 at 11:18
  • 1
    $\begingroup$ In my comment replace "full basis" in the first sentence with "full orthonormal basis". $\endgroup$ – DisintegratingByParts Apr 17 '17 at 11:29
  • $\begingroup$ It can be a orthonormal basis of a subspace rather than the full space. $\endgroup$ – mathreadler Apr 17 '17 at 11:49

It's like the Pythagorean theorem, but we might not have enough vectors in our orthonormal system to represent $x$ exactly. If $x \in \mathbb R^3$ and we have an orthonormal set of vectors $\{\phi_1,\phi_2,\phi_3\} \subset \mathbb R^3$, then we can express $x$ as a linear combination $x = \sum_{j=1}^3 \langle x,\phi_j\rangle \phi_j$. The Pythagorean theorem then tells us that $\|x\|^2=\sum_{j=1}^3 | \langle x,\phi_j\rangle |^2$. If we had only two vectors in our orthonormal set, we might not be able to represent $x$ exactly, and we would get an inequality rather than an equality.


A basis for a vector space is a set of vectors which you can construct unique linear combinations of any vector in that space:

$$\text{if } {\bf u} \in V = \{{\bf v_1,v_2,\cdots,v_n}\}$$ $$\text{then there exist unique } c_k \text{ so that } {\bf u} = \sum_{k=1}^{n} c_k{\bf v}_k$$

A basis can span a subspace of a larger vector space, if there exist vectors in the larger vector space which can not be represented as such a combination.

The inequality then happens if the basis is for a subspace which the vector we sum scalar products over does not lie in the subspace, but has components outside of it. Those components will then be kind of "invisible" to the basis.


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.