So I have a question that says,

Give a basis for the orthogonal complement of each of the following subspaces of $\mathbb{R}^4$

$w=\left \{ \boldsymbol{x} \in\mathbb{R}^4:x_1+3x_3+4x_4=0, x_2+2x_3-5x_4=0 \right \}$

I thought to find a basis for the orthogonal complement when the subspace is given you just solve for the nullspace of the matrix. However, the answer is suppose to be

{$ (1,0,3,4),(0,1,2,-5)$}.

But if I solve for the nullspace I get

{ ${(-3,2,1,0)(-4,5,0,1)}$}

So I'm confused as to what I'm doing wrong? Could someone explain?

  • $\begingroup$ Nullspace of which matrix? What it looks like is that you found basis for the original subspace, not its orthogonal complement. $\endgroup$ – Ennar Dec 7 '16 at 5:52
  • $\begingroup$ Wouldn't you just create a matrix from $ x_1+3x_3+4x_4=0,x_2+2x_3−5x_4=0$? $\endgroup$ – user6259845 Dec 7 '16 at 5:54
  • $\begingroup$ Nullspace of that matrix is the original subspace, since it is given by homongeneous equations, and not its orthogonal complement. Hint: youw equations can be written as $\langle (1,0,3,4),x\rangle = 0$ and $\langle (0,1,2,-5),x\rangle = 0$ where $x$ is in the subspace. $\endgroup$ – Ennar Dec 7 '16 at 5:58
  • $\begingroup$ Yes, I originally saw if you just take out the x from the equations you get (1,0,3,4) and (0,1,2,-5), but I was more wondering is that how your suppose to solve it? Because just taking out the x doesn't seem like a correct process to solving it. $\endgroup$ – user6259845 Dec 7 '16 at 6:05
  • $\begingroup$ The two given equations can be written as $(1,0,3,4)\cdot(x_1,x_2,x_3,x_4)=0$ and $(0,1,2,-5)\cdot(x_1,x_2,x_3,x_4)=0$. What do you know about pairs of vectors whose dot product vanishes? $\endgroup$ – amd Dec 7 '16 at 8:35

Let $S =\{x\in \Bbb R^4\mid x_1+3x_3+4x_4=0, x_2+2x_3-5x_4=0\}$. Obviously, $S$ is space of solutions of linear equation $Ax = 0$ where, $$A = \begin{pmatrix}1& 0& 3& 4\\ 0& 1& 2& -5\end{pmatrix}$$

Here comes your error, you solved equation $Ax = 0$ and found vectors $(-3,2,1,0)$ and $(-4,5,0,1)$ (which is wrong, by the way, but that's not important), but you incorrectly interpret that the solution space is orthogonal to $S$. On the contrary, the solution space is equal to $S$, as I've already said.

So, how do you find $S^\perp$? Well, by definition $S^\perp = \{x\in\Bbb R^4\mid \langle x,y\rangle = 0, \forall y\in S\}$.

Let us denote $u = (1,0,3,4)$, $v = (0,1,2,-5)$, $S' =\operatorname{span}\{u,v\}$. Then we can rewrite $$S = \{x\in\Bbb R^4\mid \langle u,x\rangle = 0, \langle v, x\rangle = 0 \} \stackrel{(*)}{=}\{x\in\Bbb R^4\mid \langle x,y\rangle = 0,\forall y\in S'\} = (S')^\perp$$ and hence, $S^\perp = ((S')^\perp)^\perp = S'$.

$(*)$ This is because of bilinearity of inner product.


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.