<br> 设A的n个两两正交的特征向量为<ruby>α<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>1</sub>,<ruby>α<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>2</sub>,…,<ruby>α<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>n</sub>,其对应的特征值依次为λ<sub>1</sub>,λ<sub>2</sub>,…,λ<sub>n</sub>。<br> 令<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>i</sub>=<ruby>α<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>i</sub>/,<ruby>α<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>i</sub>,(i=1,2,…,n),则<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>1</sub>,<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>2</sub>,…,<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>n</sub>是两两正交的单位向量。<br> 记P=(<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>1</sub>,<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>2</sub>,…,<ruby>ξ<rp>(</rp><rt style="font-size:7px;layout-grid-mode:line">→</rt><rp>)</rp></ruby><sub>n</sub>),即P是正交矩阵。从而有P<sup>-</sup><sup>1</sup>=P<sup>T</sup>,P<sup>-</sup><sup>1</sup>AP=diag(λ<sub>1</sub>,λ<sub>2</sub>,…,λ<sub>n</sub>)=Λ,即A=PΛP<sup>-</sup><sup>1</sup>=PΛP<sup>T</sup>,故A<sup>T</sup>=(PΛP<sup>T</sup>)<sup>T</sup>=(P<sup>T</sup>)<sup>T</sup>Λ<sup>T</sup>P<sup>T</sup>=PΛP<sup>T</sup>=A,即A是对称矩阵。