-
Notifications
You must be signed in to change notification settings - Fork 2
/
44A35-YoungsTheorem.tex
218 lines (198 loc) · 5.61 KB
/
44A35-YoungsTheorem.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
\documentclass[12pt]{article}
\usepackage{pmmeta}
\pmcanonicalname{YoungsTheorem}
\pmcreated{2013-03-22 18:17:44}
\pmmodified{2013-03-22 18:17:44}
\pmowner{Ziosilvio}{18733}
\pmmodifier{Ziosilvio}{18733}
\pmtitle{Young's theorem}
\pmrecord{15}{40912}
\pmprivacy{1}
\pmauthor{Ziosilvio}{18733}
\pmtype{Theorem}
\pmcomment{trigger rebuild}
\pmclassification{msc}{44A35}
\endmetadata
% this is the default PlanetMath preamble. as your knowledge
% of TeX increases, you will probably want to edit this, but
% it should be fine as is for beginners.
% almost certainly you want these
\usepackage{amssymb}
\usepackage{amsmath}
\usepackage{amsfonts}
% used for TeXing text within eps files
%\usepackage{psfrag}
% need this for including graphics (\includegraphics)
%\usepackage{graphicx}
% for neatly defining theorems and propositions
%\usepackage{amsthm}
% making logically defined graphics
%%%\usepackage{xypic}
% there are many more packages, add them here as you need them
% define commands here
\begin{document}
%% Last modified: August 10, 2011
%% MSC: 44A35
\newcommand{\ie}{\textit{i.e.}}
\newcommand{\Rset}{\ensuremath{\mathbb{R}}}
\newtheorem{theorem*}{Theorem}
Let $f,g:\Rset^n\to\Rset$.
Recall that the \emph{convolution} of $f$ and $g$ at $x$ is
\begin{displaymath}
(f\ast g)(x)=\int_{\Rset^n}f(x-y)g(y)dy
\end{displaymath}
provided the integral is defined.
The following result is due to William Henry Young.
\begin{theorem*}
Let $p,q,r\in[1,\infty]$ satisfy
\begin{equation} \label{eq:young-coeff}
\frac{1}{p}+\frac{1}{q}-\frac{1}{r}=1
\end{equation}
with the convention $1/\infty=0$.
Let $f\in L^p(\Rset^n)$, $g\in L^q(\Rset^n)$.
Then:
\begin{enumerate}
\item \label{it:L1}
The function $y\mapsto f(x-y)g(y)$
belongs to $L^1(\Rset^n)$ for almost all $x$.
\item \label{it:Lr}
The function $x\mapsto(f\ast g)(x)$
belongs to $L^r(\Rset^n)$.
\item \label{it:cpq}
There exists a constant $c=c_{p,q}\leq 1$,
depending on $p$ and $q$ but not on $f$ or $g$, such that
%%\begin{equation} \label{eq:young-conv}
\begin{displaymath}
\|{f\ast g}\|_r
\leq c\cdot\|f\|_p\cdot\|g\|_q
\end{displaymath}
%%\end{equation}
\end{enumerate}
\end{theorem*}
Observe the analogy with the similar result
with convolution replaced by ordinary (pointwise) product,
where the requirement is $1/p+1/q=1/r$---\ie,
$1/p+1/q-1/r=0$---instead of (\ref{eq:young-coeff}).
The cases
\begin{enumerate}
\item \label{item:p-q-infty}
$1/p+1/q=1$, $r=\infty$
\item \label{item:1-q-q}
$p=1$, $q \in [1,\infty)$, $r=q$
\end{enumerate}
are the most widely known;
for these we provide a proof, supposing $c_{p,q}=1$.
We shall use the following facts:
\begin{itemize}
\item
If $x\mapsto f(x),x\mapsto g(x)$ are measurable,
then $(x,y)\mapsto f(x-y)g(y)$ is measurable.
\item
For any $x$, if $f\in L^p$,
then $y\mapsto f(x-y)$ belongs to $L^p$ as well,
and its $L^p$-norm is the same as $f$'s.
\item
For any $y$, if $f\in L^p$,
then $x\mapsto f(x-y)$ belongs to $L^p$ as well,
and its $L^p$-norm is the same as $f$'s.
\end{itemize}
\textit{Proof of case~\ref{item:p-q-infty}}.
Suppose $f\in L^p(\Rset^n)$, $g\in L^q(\Rset^n)$ with $1/p+1/q=1$.
Then
\begin{displaymath}
\left|\int f(x-y)g(y)dy\right|
\leq\int|f(x-y)g(y)|dy
\leq\|f\|_p\|g\|_q\;.
\end{displaymath}
This holds for all $x\in\Rset^n$,
therefore
\begin{math}
\|f\ast g\|_\infty\leq\|f\|_p\|g\|_q
\end{math}
as well.
\textit{Proof of case~\ref{item:1-q-q}}.
First, suppose $q=1$.
We may suppose $f$ and $g$ are Borel measurable:
if they are not, we replace them with Borel measurable functions
$\tilde{f}$ and $\tilde{g}$
which are equal to $f$ and $g$, respectively,
outside of a set of Lebesgue measure zero;
apply the theorem to $\tilde{f}$, $\tilde{g}$, and $\tilde{f}\ast\tilde{g}$;
and deduce the theorem for $f$, $g$, and $f\ast g$.
By Tonelli's theorem,
\begin{displaymath}
\int\left(\int|f(x-y)g(y)|dy\right)dx
=\int\left(\int|f(x-y)|dx\right)|g(y)|dy
=\|f\|_1\|g\|_1\,,
\end{displaymath}
thus the function $(x,y)\mapsto f(x-y)g(y)$
belongs to $L^1(\Rset^n\times\Rset^n)$.
By Fubini's theorem,
the function $y\mapsto f(x-y)g(y)$
belongs to $L^1(\Rset^n)$ for almost all $x$,
and $x\mapsto(f\ast g)(x)$ belongs to $L^1(\Rset^n)$;
plus,
\begin{displaymath}
\|f\ast g\|_1
\leq\int\int|f(x-y)g(y)|dydx
=\|f\|_1\|g\|_1\;.
\end{displaymath}
Suppose now $q>1$;
choose $q'$ so that $1/q + 1/q' = 1$.
By the argument above,
\begin{math}
y \mapsto |f(x-y)| \cdot |g(y)|^q
\end{math}
belongs to $L^1$ for almost all $x$:
for those $x$, put
\begin{math}
u(y) = |f(x-y)|^{1/q'},
v(y) = |f(x-y)|^{1/q}|g(y)|.
\end{math}
Then $u \in L^{q'}$ and $v \in L^q$ with $1/q' + 1/q = 1$,
so $uv \in L^1$ and
\begin{math}
\|uv\|_1 \leq \|u\|_{q'}\|v\|_q :
\end{math}
but $uv = |f(x-y)g(y)|$, so point~\ref{it:L1} of the theorem is proved.
By H\"older's inequality,
\begin{displaymath}
\left| \int f(x-y)g(y)dy \right|
\leq
\int |f(x-y)g(y)| dy
\leq
\|f\|_1^{1/q'} \left( \int |f(x-y)| \cdot |g(y)|^q dy \right)^{1/q}
\;:
\end{displaymath}
but we know that $|f| \ast |g|^q \in L^1$,
so $f \ast g \in L^q$ and point~\ref{it:Lr} is also proved.
Finally,
\begin{displaymath}
\|f \ast g\|_q^q
\leq
\|f\|_1^{q/q'} \| |f| \ast |g|^q \|_1
\leq
\|f\|_1^{q/q'} \| f \|_1 \| g \|_q^q
=
\|f\|_1^{1+q/q'} \|g\|_q^q
\;:
\end{displaymath}
but $1/q + 1/q' = 1$ means $q+q' = qq'$ and thus $1+q/q' = q$,
so that point~\ref{it:cpq} is also proved.
\begin{thebibliography}{99}
\bibitem{gilardi}
G. Gilardi.
\textit{Analisi tre.}
McGraw-Hill 1994.
\bibitem{rudin}
W. Rudin.
\textit{Real and complex analysis.}
McGraw-Hill 1987.
\bibitem{young1912}
W. H. Young.
On the multiplication of successions of Fourier constants.
\textit{Proc. Roy. Soc. Lond. Series A} \textbf{87} (1912) 331--339.
\end{thebibliography}
%%%%%
%%%%%
\end{document}