logo

Translation in L2 Spaces: Translations, Modulations, Dilations 📂Lebesgue Spaces

Translation in L2 Spaces: Translations, Modulations, Dilations

Definition1

  • Translation is defined as Ta:L2L2T_{a} : L^{2} \to L^{2} for aRa \in \mathbb{R} as follows.

(Taf)(x):=f(xa) \left( T_{a} f \right) (x) := f(x-a)

  • Modulation is defined as Eb:L2L2E_{b} : L^{2} \to L^{2} for bRb \in \mathbb{R} as follows.

(Ebf)(x):=e2πibxf(x) \left( E_{b} f \right) (x) := e^{2 \pi i b x} f(x)

  • Dilation is defined as Dc:L2L2D_{c} : L^{2} \to L^{2} for c>0c > 0 as follows.

(Dcf)(x):=1cf(xc) \left( D_{c} f \right) (x) := {{ 1 } \over { \sqrt{c} }} f \left( {{ x } \over { c }} \right)

Explanation

The linear operators mentioned above are commonly used in the space of L2L^{2}. Although in Korean they are translated to translation, modulation, and dilation respectively, it might be easier to understand them directly in English due to the mathematical context.

The e2πibxe^{2 \pi i b x} multiplied in modulation is literally an abstracted rotation.

The 1c\displaystyle {{ 1 } \over { \sqrt{c} }} multiplied in dilation can be seen as being squared to match the norm 2\left\| \cdot \right\|_{2}. Especially for c=1/2c = 1/2, DD defined as follows plays a special role.

(Df)(x):=2f(2x) ( D f ) (x) := \sqrt{2} f (2x)

DD is written as follows for convenience with respect to jZj \in \mathbb{Z}.

(Djf)(x):=2jf(2jx) ( D^{j} f ) (x) := \sqrt{2}^{j} f \left( 2^{j} x \right)

Properties

For all a,bRa, b \in \mathbb{R}, c>0c > 0, and f,gL1f,g \in L^{1},

  1. Ta,Eb,DcT_{a} , E_{b}, D_{c} is a bounded linear operator.

  2. Inverse operator: Ta,Eb,DcT_{a} , E_{b}, D_{c} is unitary.

  3. Commutation relations:

(TaEbf)(x)=e2πiba(EbTaf)(x)(TaDcf)(x)=(DcTa/cf)(x)(DcEbf)(x)=(Eb/cDcf)(x) (T_{a} E_{b} f ) (x) = e^{- 2 \pi i b a} (E_{b} T_{a} f ) (x) \\ (T_{a} D_{c} f ) (x) = (D_{c} T_{a/c} f ) (x) \\ (D_{c} E_{b} f ) (x) = (E_{b/c} D_{c} f ) (x)

  • Relation with Fourier transform:

    FTa=EaFFEb=TbFFDc=D1/cF \mathcal{F} T_{a} = E_{-a} \mathcal{F} \\ \mathcal{F} E_{b} = T_{b} \mathcal{F} \\ \mathcal{F} D_{c} = D_{1/c} \mathcal{F}

    Regarding DD, the following can be obtained about j,kZj, k \in \mathbb{Z} as corollaries of the above theorems.

    TkDj=DjT2jkDjTk=T2jkDj(Dj)=Dj T_{k} D^{j} = D^{j} T_{2^{j} k } \\ D^{j} T_{k} = T_{2^{-j}k} D^{j} \\ \left( D^{j} \right)^{ \ast } = D^{-j}

Proof

1.

  • Part 1. Linear

    For all f,gL2f,g \in L^{2} and α,βC\alpha , \beta \in \mathbb{C},

    Ta(αf+βg)(x)=(αf+βg)(xa)=αf(xa)+βg(xa)=αTaf(x)+βTag(x) \begin{align*} T_{a} \left( \alpha f + \beta g \right)(x) =& \left( \alpha f + \beta g \right)(x-a) \\ =& \alpha f (x-a) + \beta g (x-a) \\ =& \alpha T_{a} f (x) + \beta T_{a} g (x) \end{align*}

    Therefore, TaT_{a} is linear.

    Eb(αf+βg)(x)=e2πibx(αf+βg)(x)=αe2πibxf(x)+βe2πibxg(x)=αEbf(x)+βEbg(x) \begin{align*} E_{b} \left( \alpha f + \beta g \right)(x) =& e^{ 2 \pi i b x } \left( \alpha f + \beta g \right)(x) \\ =& \alpha e^{ 2 \pi i b x } f (x) + \beta e^{ 2 \pi i b x } g (x) \\ =& \alpha E_{b} f (x) + \beta E_{b} g (x) \end{align*}

    Therefore, EbE_{b} is linear.

    Dc(αf+βg)(x)=1c(αf+βg)(xc)=α1cf(x)+β1cg(x)=αDcf(x)+βDcg(x) \begin{align*} D_{c} \left( \alpha f + \beta g \right)(x) =& {{ 1 } \over { \sqrt{c} }} \left( \alpha f + \beta g \right) \left( {{ x } \over { c }} \right) \\ =& \alpha {{ 1 } \over { \sqrt{c} }} f (x) + \beta {{ 1 } \over { \sqrt{c} }} g (x) \\ =& \alpha D_{c} f (x) + \beta D_{c} g (x) \end{align*}

    Therefore, DcD_{c} is linear.

  • Part 2. Bounded

    Substitute as t:=xat := x - a,

    Taf2=Taf(x)2dx=f(xa)2dx=f(t)2dt=f2 \begin{align*} \left\| T_{a} f \right\|_{2} =& \int_{-\infty}^{\infty} \left| T_{a} f \left( x \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} \left| f \left( x - a \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} \left| f \left( t \right) \right|^{2} dt \\ =& \left\| f \right\|_{2} \end{align*}

    Therefore, TaT_{a} is bounded. Since e2πibx=1\left| e^{2 \pi i b x } \right| =1,

    Ebf2=Ebf(x)2dx=e2πibxf(x)2dx=1f(t)2dt=f2 \begin{align*} \left\| E_{b} f \right\|_{2} =& \int_{-\infty}^{\infty} \left| E_{b} f \left( x \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} \left| e^{2 \pi i b x } f \left( x \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} 1 \cdot \left| f \left( t \right) \right|^{2} dt \\ =& \left\| f \right\|_{2} \end{align*}

    Therefore, EbE_{b} is bounded. Substitute as t:=x/ct := x/c,

    Dcf2=Dcf(x)2dx=1cf(xc)2dx=1cf(t)2cdt=f(t)2dt=f2 \begin{align*} \left\| D_{c} f \right\|_{2} =& \int_{-\infty}^{\infty} \left| D_{c} f \left( x \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} \left| {{ 1 } \over { \sqrt{c} }} f \left( {{ x } \over { c }} \right) \right|^{2} dx \\ =& \int_{-\infty}^{\infty} {{ 1 } \over { c }} \left| f \left( t \right) \right|^{2} c dt \\ =& \int_{-\infty}^{\infty} \left| f \left( t \right) \right|^{2} dt \\ =& \left\| f \right\|_{2} \end{align*}

    Therefore, DcD_{c} is bounded.


  1. Ole Christensen, Functions, Spaces, and Expansions: Mathematical Tools in Physics and Engineering (2010), p120-122 ↩︎