Problem Statement: Let $A=\begin{bmatrix} 2 && 1 \\ 1 && 2 \end{bmatrix}$. Find an orthonormal basis for $\mathbb{C}^2$ with respect to the Hermitian form $\bar{x}^TAy$.

I am trying to figure out this proof, but not sure if what I am currently trying to do will help me. I was goin to start out by taking an arbitrary basis $B=\left\{v_1,v_2\right\}$ and write two arbitrary vectors in $\mathbb{C}^2$ as linear combinations of basis elements: $$v=x_1v_1+x_2v_2$$ $$w=y_1v_1+y_2v_2$$

Then break down the Hermitian form $$\langle v,w \rangle=\langle x_1v_1+x_2v_2, y_1v_1+y_2v_2 \rangle=\bar{x_1}\langle v_1, y_1v_1+y_2v_2 \rangle+\bar{x_2}\langle v_2, y_1v_1+y_2v_2 \rangle=\bar{x_1}y_1\langle v_1, v_1 \rangle+\bar{x_1}y_2\langle v_1, v_2 \rangle+\bar{x_2}y_1\langle v_2, v_1 \rangle+\bar{x_2}y_2\langle v_2, v_2 \rangle=\sum_{i=1,j=1}^2 \bar{x}_iy_j\langle v_i, v_j\rangle=\bar{x}^TAy$$

Then I want to use the fact that we're given $A$ in order to define $v_1,v_2$, but I am confused with one thing: In my notes, I have that $$a_{ij}=\langle v_i,v_j\rangle$$ but my notes also say for any orthonormal basis, that $$\langle v_i,v_j\rangle:=\begin{cases} 1\ \mathrm{if}\ i=j \\ 0\ \mathrm{if}\ i\neq j \end{cases}$$ which conflicts with the fact that $A=\begin{bmatrix} 2 && 1 \\ 1 && 2 \end{bmatrix}$ because that would mean that $$\langle v_i,v_j\rangle:=\begin{cases} 2\ \mathrm{if}\ i=j \\ 1\ \mathrm{if}\ i\neq j \end{cases}$$

Is there a more straightforward approach to this problem?


The straight-forward approach is Gram-Schmidt.

Start with the standard (canonical) basis $\{e_1,e_2\}$. Our goal is to get an orthonormal basis $\{v_1,v_2\}$.

First, we take $$ v_1 = \frac {1}{\sqrt{\langle e_1,e_1 \rangle}} e_1 = \frac 1 {\sqrt 2}e_1 $$ Now, take $$ w_2 = e_2 - \langle e_2,v_1 \rangle v_1 = e_2 - \frac 1 2 \langle e_2,e_1 \rangle e_1 = e_2 - \frac 12 e_1 $$ Noting that $$ \langle w_2,w_2\rangle = \frac 12 - \frac 12 - \frac 12 + 2 = \frac 32 $$ and set $$ v_2 = \frac{1}{\sqrt{\langle w_2,w_2 \rangle}}w_2 = \sqrt{\frac{2}{3}}(e_2 - \frac 12 e_1) $$ so, now we have an orthonormal basis.

  • $\begingroup$ I am confused how this is with repsect to the Hermitian form $\bar{x}^TAy$. Why did we select the basis that we did? It seems that the matrix $A$ was pointless in this case? or did we use the fact that $a_{11}=2$ to define $\langle e_1,e_1 \rangle=2$ and $a_{21}=1$ to define $\langle e_2,e_1 \rangle=1$? $\endgroup$ – yung_Pabs Feb 7 '16 at 0:38
  • 1
    $\begingroup$ I didn't define $\langle e_1,e_1 \rangle$, I computed it using the definition from your question (without showing my work). Check that I did so correctly. By the "standard basis", I mean specifically that $$ e_1 = \pmatrix{1\\0}, \quad e_2 = \pmatrix{0\\1} $$ $\endgroup$ – Omnomnomnom Feb 7 '16 at 0:43
  • $\begingroup$ So are you computing the inner product with $a_{ij}=⟨v_i,v_j⟩$ using our given matrix $A$? $\endgroup$ – yung_Pabs Feb 7 '16 at 0:47
  • 1
    $\begingroup$ Yes!${}{}{}{}{}{}{}{}{}$ $\endgroup$ – Omnomnomnom Feb 7 '16 at 0:50

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.