Linear Algebra Home

Orthogonal Vectors in \(\mathbb R^n\)

    


Definition. The inner product or the dot product of two vectors \(\overrightarrow{u}\) and \(\overrightarrow{v}\) in \(\mathbb R^n\), denoted by \(\overrightarrow{u} \cdot \overrightarrow{v}\), is defined by \(\overrightarrow{u} \cdot \overrightarrow{v}=\overrightarrow{u}^T \overrightarrow{v}\).

Example. For \(\overrightarrow{u}=\left[\begin{array}{r}1\\-2\\3\end{array} \right]\) and \(\overrightarrow{v}=\left[\begin{array}{r}2\\1\\-1\end{array} \right]\), \(\overrightarrow{u} \cdot \overrightarrow{v}=\overrightarrow{u}^T \overrightarrow{v}=1\cdot 2-2\cdot 1+3\cdot(-1)=-3\).

Theorem. The following are true for all \(\overrightarrow{u}\), \(\overrightarrow{v}\), \(\overrightarrow{w}\) in \(\mathbb R^n\) and for all scalars \(c\), \(d\) in \(\mathbb R\).

  1. \(\overrightarrow{u} \cdot \overrightarrow{v}=\overrightarrow{v} \cdot \overrightarrow{u}\). (symmetry)

  2. \((c \overrightarrow{u} +d \overrightarrow{v})\cdot w=c(\overrightarrow{u} \cdot \overrightarrow{w})+d (\overrightarrow{v} \cdot \overrightarrow{w})\). (linearity)

  3. \(\overrightarrow{u} \cdot \overrightarrow{u}\geq 0\) where \(\overrightarrow{u} \cdot \overrightarrow{u}=0\) if and only if \(\overrightarrow{u}=\overrightarrow{0}\). (nonnegativity)

Definition. The length or norm of \(\overrightarrow{v}=[v_1,v_2,\ldots,v_n]^T\) in \(\mathbb R^n\), denoted by \(\left\lVert\overrightarrow{v}\right\rVert\), is defined by \(\left\lVert\overrightarrow{v}\right\rVert =\sqrt{v_1^2+v_2^2+\cdots+v_n^2}\). \(\overrightarrow{v}\in \mathbb R^n\) is a unit vector if \(\left\lVert\overrightarrow{v}\right\rVert=1\).

Remark. The following are true for all \(\overrightarrow{v}\) in \(\mathbb R^n\) and for all scalars \(c\) in \(\mathbb R\).

  1. \(\left\lVert\overrightarrow{v}\right\rVert^2=\overrightarrow{v} \cdot \overrightarrow{v}\).

  2. \(\left\lVert c\overrightarrow{v}\right\rVert=|c| \left\lVert\overrightarrow{v}\right\rVert\).

  3. The unit vector in the direction of \(\overrightarrow{v}\neq \overrightarrow{0}\) is \(\frac{1}{\left\lVert\overrightarrow{v}\right\rVert}\overrightarrow{v}\).

Example. The unit vector in the opposite direction of \(\overrightarrow{v}=\left[\begin{array}{r}1\\-2\\3\end{array} \right]\) is \(\frac{-1}{\left\lVert\overrightarrow{v}\right\rVert}\overrightarrow{v} =\frac{1}{\sqrt{14}} \left[\begin{array}{r}-1\\2\\-3\end{array} \right]\).

Definition. The distance between \(\overrightarrow{u},\overrightarrow{v}\) in \(\mathbb R^n\), denoted by \(\operatorname{d} (\overrightarrow{u},\overrightarrow{v})\), is defined by \[\operatorname{d}(\overrightarrow{u},\overrightarrow{v})=\left\lVert\overrightarrow{u}-\overrightarrow{v}\right\rVert.\]
Note that \(\operatorname{d}(\overrightarrow{u},\overrightarrow{v})^2=\left\lVert\overrightarrow{u}-\overrightarrow{v}\right\rVert^2 =\left\lVert\overrightarrow{u}\right\rVert^2+\left\lVert\overrightarrow{v}\right\rVert^2-2 \overrightarrow{u} \cdot \overrightarrow{v}\) and \(\operatorname{d}(\overrightarrow{u},-\overrightarrow{v})^2=\left\lVert\overrightarrow{u}+\overrightarrow{v}\right\rVert^2 =\left\lVert\overrightarrow{u}\right\rVert^2+\left\lVert\overrightarrow{v}\right\rVert^2+2 \overrightarrow{u} \cdot \overrightarrow{v}\). So \(\overrightarrow{u}\) and \(\overrightarrow{v}\) are perpendicular if and only if \(\operatorname{d}(\overrightarrow{u},\overrightarrow{v}) =\operatorname{d}(\overrightarrow{u},-\overrightarrow{v})\) if and only if \(\overrightarrow{u} \cdot \overrightarrow{v}=0\).

Definition. Two vectors \(\overrightarrow{u}\) and \(\overrightarrow{v}\) in \(\mathbb R^n\) are orthogonal if \(\overrightarrow{u} \cdot \overrightarrow{v}=0\).

Example. Let \(\overrightarrow{u}=[3,2,-5,0]^T\) and \(\overrightarrow{v}=[-4,1,-2,1]^T\).

  1. Determine if \(\overrightarrow{u}\) and \(\overrightarrow{v}\) are orthogonal.

  2. Find \(\operatorname{d}(\overrightarrow{u},\overrightarrow{v})\).

Solution. (a) Since \(\overrightarrow{u} \cdot \overrightarrow{v}=3\cdot(-4)+2\cdot1-5\cdot(-2)+0\cdot 1=0\), \(\overrightarrow{u}\) and \(\overrightarrow{v}\) are orthogonal.

(b) \[\begin{align*} \operatorname{d}(\overrightarrow{u},\overrightarrow{v}) =\left\lVert\overrightarrow{u}-\overrightarrow{v}\right\rVert &=\sqrt{\left\lVert\overrightarrow{u}\right\rVert^2+\left\lVert\overrightarrow{v}\right\rVert^2 -2 \overrightarrow{u} \cdot \overrightarrow{v} }\\ &=\sqrt{\left\lVert\overrightarrow{u}\right\rVert^2+\left\lVert\overrightarrow{v}\right\rVert^2} \;(\text{since } \overrightarrow{u} \cdot \overrightarrow{v}=0)\\ &=\sqrt{38+22}\\ &=\sqrt{60} \end{align*}\]

Theorem.(Pythagorean Theorem) Two vectors \(\overrightarrow{u}\) and \(\overrightarrow{v}\) in \(\mathbb R^n\) are orthogonal if and only if \(\left\lVert\overrightarrow{u}+\overrightarrow{v}\right\rVert^2=\left\lVert\overrightarrow{u}\right\rVert^2 +\left\lVert\overrightarrow{v}\right\rVert^2\).

Note that \[\left\lVert\overrightarrow{u}+\overrightarrow{v}\right\rVert^2=(\overrightarrow{u}+\overrightarrow{v})\cdot (\overrightarrow{u}+\overrightarrow{v})=\overrightarrow{u}\cdot \overrightarrow{u}+\overrightarrow{u}\cdot \overrightarrow{u}+2\overrightarrow{u}\cdot \overrightarrow{v} =\left\lVert\overrightarrow{u}\right\rVert^2+\left\lVert\overrightarrow{v}\right\rVert^2+2\overrightarrow{u}\cdot \overrightarrow{v}.\] Then \(\left\lVert\overrightarrow{u}+\overrightarrow{v}\right\rVert^2=\left\lVert\overrightarrow{u}\right\rVert^2 +\left\lVert\overrightarrow{v}\right\rVert^2\) if and only if \(\overrightarrow{u}\cdot \overrightarrow{v}=0\).

Definition. The angle \(\theta\) between two vectors \(\overrightarrow{u}\) and \(\overrightarrow{v}\) in \(\mathbb R^n\) is the angle in \([0,\pi]\) satisfying \[\overrightarrow{u} \cdot \overrightarrow{v}=\left\lVert\overrightarrow{u}\right\rVert \left\lVert\overrightarrow{v}\right\rVert \cos \theta.\]
Definition. Let \(W\) be a subspace of \(\mathbb R^n\). A vector \(\overrightarrow{v}\in \mathbb R^n\) is orthogonal to \(W\) if \(\overrightarrow{v}\cdot \overrightarrow{w} =0\) for all \(\overrightarrow{w}\in W\). The orthogonal complement of \(W\), denoted by \(W^{\perp}\), is the set of all vectors in \(\mathbb R^n\) that are orthogonal to \(W\), i.e., \[W^{\perp}=\{\overrightarrow{v}\in \mathbb R^n \;|\; \overrightarrow{v}\cdot \overrightarrow{w} =0 \text{ for all } \overrightarrow{w}\in W\}.\]
Example.

  1. If \(L\) is a line in \(\mathbb R^2\) through the origin, then \(L^{\perp}\) is the line through the origin that is perpendicular to \(L\).

  2. If \(L\) is a line in \(\mathbb R^3\) through the origin, then \(L^{\perp}\) is the plane through the origin that is perpendicular to \(L\). Note that \((L^{\perp})^{\perp}=L\).

Theorem. Let \(W\) be a subspace of \(\mathbb R^n\) and \(W=\operatorname{Span} \{\overrightarrow{w_1},\overrightarrow{w_2},\ldots,\overrightarrow{w_k}\}\). Then

  1. \(\overrightarrow{v} \in W^{\perp}\) if and only if \(\overrightarrow{v}\cdot \overrightarrow{w_i}=0\) for \(i=1,2,\ldots,k\).

  2. \(W^{\perp}\) is a subspace of \(\mathbb R^n\).

  3. \((W^{\perp})^{\perp}=W\).

  4. \(W\cap W^{\perp}=\{\overrightarrow{0}\}\).
  1. Let \(\overrightarrow{v} \in W^{\perp}\). Then \(\overrightarrow{v}\cdot \overrightarrow{w} =0\) for all \(\overrightarrow{w}\in W\). Since \(\overrightarrow{w_i}\in W\) for \(i=1,2,\ldots,k\), \(\overrightarrow{v}\cdot \overrightarrow{w_i} =0\) for \(i=1,2,\ldots,k\). Conversely suppose that \(\overrightarrow{v}\cdot \overrightarrow{w_i}=0\) for \(i=1,2,\ldots,k\). Let \(\overrightarrow{w}\in W=\operatorname{Span} \{\overrightarrow{w_1},\overrightarrow{w_2},\ldots,\overrightarrow{w_k}\}\). Then \(\overrightarrow{w}=c_1\overrightarrow{w_1} +c_2\overrightarrow{w_2}+\cdots+c_k\overrightarrow{w_k}\) for some scalars \(c_1,c_2,\ldots,c_k\). Then \[\overrightarrow{v}\cdot \overrightarrow{w}= \overrightarrow{v}\cdot (c_1\overrightarrow{w_1} +c_2\overrightarrow{w_2}+\cdots+c_k\overrightarrow{w_k}) =c_1(\overrightarrow{v}\cdot\overrightarrow{w_1}) +c_2(\overrightarrow{v}\cdot\overrightarrow{w_2})+\cdots+c_k(\overrightarrow{v}\cdot\overrightarrow{w_k})=0.\] Thus \(\overrightarrow{v}\cdot \overrightarrow{w} =0\) for all \(\overrightarrow{w}\in W\) and consequently \(\overrightarrow{v} \in W^{\perp}\).

  2. \(\overrightarrow{0}\cdot \overrightarrow{w} =0\) for all \(\overrightarrow{w}\in W\), \(\overrightarrow{0} \in W^{\perp}\) and \(W^{\perp}\neq \varnothing\). Let \(\overrightarrow{u},\overrightarrow{v}\in W^{\perp}\) and \(c,d\in \mathbb R\). Then for all \(\overrightarrow{w}\in W\), \[(c\overrightarrow{u}+d\overrightarrow{v})\cdot \overrightarrow{w}= c(\overrightarrow{u}\cdot \overrightarrow{w})+d(\overrightarrow{v}\cdot \overrightarrow{w})=c\overrightarrow{0}+d\overrightarrow{0}=\overrightarrow{0}.\] Thus \(c\overrightarrow{u}+d\overrightarrow{v} \in W^{\perp}\). Therefore \(W^{\perp}\) is a subspace of \(\mathbb R^n\).

  3. Exercise.

  4. First note that \(\{\overrightarrow{0}\}\subseteq W\cap W^{\perp}\). Let \(\overrightarrow{v}\in W\cap W^{\perp}\). Then \(\overrightarrow{v}\in W\) and \(\overrightarrow{v}\in W^{\perp}\). Thus \(\left\lVert\overrightarrow{v}\right\rVert^2=\overrightarrow{v}\cdot \overrightarrow{v}=0\) which implies \(\overrightarrow{v}=\overrightarrow{0}\). Therefore \(W\cap W^{\perp}=\{\overrightarrow{0}\}\).

Theorem. Let \(A\) be an \(m\times n\) real matrix. Then \(\operatorname{RS}(A)^{\perp}=\operatorname{NS}(A)\) and \(\operatorname{CS}\left(A\right)^{\perp}=\operatorname{NS}(A^T)\).

To show \(\operatorname{NS}(A)\subseteq \operatorname{RS}(A)^{\perp}\), let \(\overrightarrow{x}\in \operatorname{NS}(A)=\{ \overrightarrow{x}\in \mathbb R^n \;|\; A\overrightarrow{x}=\overrightarrow{0}\}\). Then each row of \(A\) is orthogonal to \(\overrightarrow{x}\). Since \(\operatorname{RS}(A)\) is the span of rows of \(A\), \(\overrightarrow{x}\) is orthogonal to each vector of \(\operatorname{RS}(A)\). Then \(\overrightarrow{x}\in \operatorname{RS}(A)^{\perp}\). Thus \(\operatorname{NS}(A)\subseteq \operatorname{RS}(A)^{\perp}\). To show \(\operatorname{RS}(A)^{\perp}=\operatorname{NS}(A)\), it suffices to show \(\operatorname{RS}(A)^{\perp}\subseteq \operatorname{NS}(A)\). Let \(\overrightarrow{x}\in \operatorname{RS}(A)^{\perp}\). Since rows of \(A\) are in \(\operatorname{RS}(A)\), \(\overrightarrow{x}\) is orthogonal to each row of \(A\). Then \(A\overrightarrow{x}=\overrightarrow{0}\) and \(\overrightarrow{x}\in \operatorname{NS}(A)\). Thus \(\operatorname{RS}(A)^{\perp}\subseteq \operatorname{NS}(A)\).

Finally \(\operatorname{NS}(A^T)=\operatorname{RS}(A^T)^{\perp}=\operatorname{CS}\left(A\right)^{\perp}\) because \(\operatorname{RS}(A^T)=\operatorname{CS}\left(A\right)\).


Last edited