Fourier Series and Bessel's Inequality
📂Fourier Analysis Fourier Series and Bessel's Inequality If a function f f f defined on the interval [ − L , L ) [-L,L) [ − L , L ) is Riemann integrable , the following inequality holds, known as the Bessel’s inequality.
1 4 ∣ a 0 ∣ 2 + 1 2 ∑ n = 1 ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = ∑ n = − ∞ ∞ ∣ c n ∣ 2 ≤ 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t
\dfrac{1}{4}|a_{0}|^{2} +\dfrac{1}{2}\sum\limits_{n=1}^{\infty} \left(|a_{n}|^{2} + |b_{n}|^{2} \right) =\sum \limits_{n=-\infty}^{\infty} | c_{n} |^{2} \le \dfrac{1}{2L}\int_{-L}^{L} | f(t)|^{2} dt
4 1 ∣ a 0 ∣ 2 + 2 1 n = 1 ∑ ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = n = − ∞ ∑ ∞ ∣ c n ∣ 2 ≤ 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t
Here, a 0 , a n , b n a_{0},\ a_{n},\ b_{n} a 0 , a n , b n is the Fourier coefficient of f f f , and c n c_{n} c n is the complex Fourier coefficient of f f f .
Proof For any complex number z z z , since ∣ z ∣ 2 = z z ‾ |z|^{2}= z \overline{z} ∣ z ∣ 2 = z z ,
0 ≤ ∣ f ( t ) − ∑ n = − N N c n e i n π t L ∣ 2 = ( f ( t ) − ∑ n = − N N c n e i n π t L ) ( f ( t ) ‾ − ∑ n = − N N c n ‾ e − i n π t L ) = ∣ f ( t ) ∣ 2 − ∑ n = − N N ( f ( t ) c n ‾ e − i n π t L + f ( t ) ‾ c n e i n π t L ) + ∑ m = − N N ∑ n = − N N ∣ c n ∣ 2 e i ( m − n ) π t L
\begin{align*}
0 &\le \left| f(t)-\sum\limits_{n=-N}^{N}c_{n} e^{i\frac{n\pi t}{L}} \right|^{2}
\\ &=\left(f(t)-\sum\limits_{n=-N}^{N}c_{n} e^{i\frac{n\pi t}{L}} \right) \left(\overline{f(t)}-\sum\limits_{n=-N}^{N} \overline{c_{n}} e^{-i\frac{n\pi t}{L}} \right)
\\ &= |f(t)|^{2} - \sum \limits_{n=-N}^{N} \left( f(t) \overline{c_{n}} e^{-i\frac{n\pi t}{L}} + \overline{f(t)}c_{n} e^{i\frac{n\pi t}{L}} \right) + \sum \limits_{m=-N}^{N}\sum\limits_{n=-N}^{N} |c_{n}|^{2}e^{i\frac{(m-n)\pi t}{L}}
\end{align*}
0 ≤ f ( t ) − n = − N ∑ N c n e i L nπ t 2 = ( f ( t ) − n = − N ∑ N c n e i L nπ t ) ( f ( t ) − n = − N ∑ N c n e − i L nπ t ) = ∣ f ( t ) ∣ 2 − n = − N ∑ N ( f ( t ) c n e − i L nπ t + f ( t ) c n e i L nπ t ) + m = − N ∑ N n = − N ∑ N ∣ c n ∣ 2 e i L ( m − n ) π t
As can be seen from the first line, the expression is greater than or equal to 0 0 0 , hence the integrated value is also greater than or equal to 0 0 0 . Therefore, the following inequality holds.
0 ≤ 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t − ∑ n = − N N ( c n ‾ 1 2 L ∫ − L L f ( t ) e − i n π t L d t + c n 1 2 L ∫ − L L f ( t ) ‾ e i n π t L d t ) + ∑ m = − N N ∑ n = − N N ∣ c n ∣ 2 1 2 L ∫ − L L e i ( m − n ) π t L d t
\begin{align*}
0& \le \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2} dt - \sum \limits_{n=-N}^{N} \left(\overline{c_{n}} \dfrac{1}{2L}\int_{-L}^{L}f(t) e^{-i\frac{n\pi t}{L}}dt + c_{n}\dfrac{1}{2L}\int_{-L}^{L} \overline{f(t)} e^{i\frac{n\pi t}{L}}dt \right)
\\ &\quad+ \sum \limits_{m=-N}^{N}\sum\limits_{n=-N}^{N} |c_{n}|^{2}\dfrac{1}{2L}\int_{-L}^{L}e^{i\frac{(m-n)\pi t}{L}}dt
\end{align*}
0 ≤ 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t − n = − N ∑ N ( c n 2 L 1 ∫ − L L f ( t ) e − i L nπ t d t + c n 2 L 1 ∫ − L L f ( t ) e i L nπ t d t ) + m = − N ∑ N n = − N ∑ N ∣ c n ∣ 2 2 L 1 ∫ − L L e i L ( m − n ) π t d t
Using the definition of complex Fourier coefficients for the second term, and the orthogonality of the exponential function for the third term, it simplifies to the following.
0 ≤ 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t − ∑ n = − N N ( c ˉ n 1 2 L ∫ − L L f ( t ) e − i n π t L d t + c n 1 2 L ∫ − L L f ˉ ( t ) e i n π t L d t ) + ∑ m = − N N ∑ n = − N N ∣ c n ∣ 2 1 2 L ∫ − L L e i ( m − n ) π L t d t = 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t − ∑ n = − N N ( c ˉ n c n + c n c ˉ n ) + ∑ m = − N N ∑ n = − N N ∣ c n ∣ 2 δ m n = 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t − ∑ n = − N N 2 ∣ c n ∣ 2 + ∑ n = − N N ∣ c n ∣ 2 = 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t − ∑ n = − N N ∣ c n ∣ 2
\begin{align*}
0 &\le \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2} dt - \sum \limits_{n=-N}^{N} \left(\bar c_{n} \dfrac{1}{2L}\int_{-L}^{L}f(t) e^{-i\frac{n\pi t}{L}}dt + c_{n}\dfrac{1}{2L}\int_{-L}^{L} \bar f(t)e^{i\frac{n\pi t}{L}}dt \right) + \sum \limits_{m=-N}^{N}\sum\limits_{n=-N}^{N} |c_{n}|^{2}\dfrac{1}{2L}\int_{-L}^{L}e^{i\frac{(m-n)\pi}{L}t}dt
\\ &= \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2} dt - \sum \limits_{n=-N}^{N} \left(\bar c_{n} c_{n} + c_{n}\bar c_{n} \right) + \sum \limits_{m=-N}^{N}\sum\limits_{n=-N}^{N} |c_{n}|^{2}\delta_{mn}
\\ &= \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2} dt - \sum \limits_{n=-N}^{N}2 |c_{n}|^{2} + \sum \limits_{n=-N}^{N}|c_{n}|^{2}
\\ &= \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2} dt - \sum \limits_{n=-N}^{N} |c_{n}|^{2}
\end{align*}
0 ≤ 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t − n = − N ∑ N ( c ˉ n 2 L 1 ∫ − L L f ( t ) e − i L nπ t d t + c n 2 L 1 ∫ − L L f ˉ ( t ) e i L nπ t d t ) + m = − N ∑ N n = − N ∑ N ∣ c n ∣ 2 2 L 1 ∫ − L L e i L ( m − n ) π t d t = 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t − n = − N ∑ N ( c ˉ n c n + c n c ˉ n ) + m = − N ∑ N n = − N ∑ N ∣ c n ∣ 2 δ mn = 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t − n = − N ∑ N 2∣ c n ∣ 2 + n = − N ∑ N ∣ c n ∣ 2 = 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t − n = − N ∑ N ∣ c n ∣ 2
Taking the limit of N → ∞ N \rightarrow \infty N → ∞ and moving the second term on the last expression yields the following.
∑ n = − ∞ ∞ ∣ c n ∣ 2 ≤ 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t
\sum \limits_{n=-\infty}^{\infty} |c_{n}|^{2} \le \dfrac{1}{2L}\int_{-L}^{L} |f(t)|^{2}dt
n = − ∞ ∑ ∞ ∣ c n ∣ 2 ≤ 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t
According to the relationship between the Fourier coefficients and complex Fourier coefficients , the following equations are obtained.
∣ a n ∣ 2 + ∣ b n ∣ 2 = a n a ˉ n + b n b ˉ n = ( c n + c − n ) ( c ˉ n + c ˉ − n ) + i ( c n − c − n ) ( − i ) ( c ˉ n − c ˉ − n ) = 2 c n c ˉ n + 2 c − n c ˉ − n = 2 ( ∣ c n ∣ 2 + ∣ c − n ∣ 2 )
\begin{align*}
|a_{n}|^{2}+|b_{n}|^{2} &= a_{n} \bar a_{n} + b_{n} \bar b_{n}
\\ &= (c_{n}+c_{-n})(\bar c_{n} + \bar c_{-n})+ i (c_{n}-c_{-n})(-i)(\bar c_{n} - \bar c_{-n} )
\\ &= 2c_{n}\bar c_{n} + 2c_{-n}\bar c_{-n}
\\ &= 2\left( |c_{n}|^{2} + |c_{-n}|^{2} \right)
\end{align*}
∣ a n ∣ 2 + ∣ b n ∣ 2 = a n a ˉ n + b n b ˉ n = ( c n + c − n ) ( c ˉ n + c ˉ − n ) + i ( c n − c − n ) ( − i ) ( c ˉ n − c ˉ − n ) = 2 c n c ˉ n + 2 c − n c ˉ − n = 2 ( ∣ c n ∣ 2 + ∣ c − n ∣ 2 )
⟹ 1 2 ∑ n = 1 ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = ∑ n = 1 ∞ ( ∣ c n ∣ 2 + ∣ c − n ∣ 2 ) = ∑ n = − ∞ n ≠ 0 ∞ ∣ c n ∣ 2
\implies \dfrac{1}{2} \sum \limits_{n=1}^{\infty} \left( |a_{n}|^{2}+|b_{n}|^{2} \right) = \sum \limits_{n=1}^{\infty} \left ( |c_{n}|^{2} + |c_{-n}|^{2} \right) = \sum \limits_{n=-\infty \\ n\ne 0}^{\infty} | c_n |^{2}
⟹ 2 1 n = 1 ∑ ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = n = 1 ∑ ∞ ( ∣ c n ∣ 2 + ∣ c − n ∣ 2 ) = n = − ∞ n = 0 ∑ ∞ ∣ c n ∣ 2
And since ∣ c 0 ∣ 2 = 1 4 ∣ a 0 ∣ 2 |c_{0}|^{2} = \dfrac{1}{4}|a_{0}|^{2} ∣ c 0 ∣ 2 = 4 1 ∣ a 0 ∣ 2 ,
1 4 ∣ a 0 ∣ 2 + 1 2 ∑ n = 1 ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = ∑ n = − ∞ ∞ ∣ c n ∣ 2 ≤ 1 2 L ∫ − L L ∣ f ( t ) ∣ 2 d t
\dfrac{1}{4}|a_0|^{2} +\dfrac{1}{2}\sum\limits_{n=1}^{\infty} \left(|a_n|^{2} + |b_n|^{2} \right) =\sum \limits_{n=-\infty}^{\infty} | c_n |^{2} \le \dfrac{1}{2L}\int_{-L}^{L} | f(t)|^{2} dt
4 1 ∣ a 0 ∣ 2 + 2 1 n = 1 ∑ ∞ ( ∣ a n ∣ 2 + ∣ b n ∣ 2 ) = n = − ∞ ∑ ∞ ∣ c n ∣ 2 ≤ 2 L 1 ∫ − L L ∣ f ( t ) ∣ 2 d t
■