Interpolation Inequalities in Lebesgue Spaces
📂Lebesgue Spaces Interpolation Inequalities in Lebesgue Spaces Theorem Let’s say Ω ⊂ R n \Omega \subset \mathbb{R}^{n} Ω ⊂ R n is an open set . Suppose that for some θ \theta θ , 1 ≤ p < q < r ≤ ∞ 1 \le p \lt q\lt r \le \infty 1 ≤ p < q < r ≤ ∞ satisfies the following equation being 0 < θ < 1 0 \lt \theta \lt 1 0 < θ < 1 .
1 q = θ p + 1 − θ r
\dfrac{1}{q} = \frac{\theta}{p} + \frac{1-\theta}{r}
q 1 = p θ + r 1 − θ
Assume that u ∈ L p ( Ω ) ∩ L r ( Ω ) u \in L^p(\Omega) \cap L^r(\Omega) u ∈ L p ( Ω ) ∩ L r ( Ω ) . Then, u ∈ L q ( Ω ) u\in L^{q}(\Omega) u ∈ L q ( Ω ) holds, and the following inequality is established.
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ r 1 − θ
\left\| u \right\|_{q} \le \left\| u \right\|_{p}^{\theta} \left\| u \right\|_{r}^{1-\theta}
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ r 1 − θ
This is called the interpolation inequality .
Explanation Translating interpolation means interpolating , which implies the meaning of filling in the gaps.
For any p , r p, r p , r that is greater than or equal to 1, if u ∈ L p u\in L^{p} u ∈ L p and u ∈ L r u\in L^{r} u ∈ L r , then it is guaranteed that for all q q q between p p p and r r r , u ∈ L q u\in L^q u ∈ L q holds.
Proof First, by multiplying both sides of the given assumption by q q q , we have
1 = θ q p + ( 1 − θ ) q r ⟹ 1 = 1 p θ q + 1 r ( 1 − θ ) q
\begin{align*}
&& 1 =&\ \dfrac{ \theta q}{p}+\dfrac{(1-\theta) q}{r}
\\ \implies && 1 =&\ \dfrac{1}{\frac{p}{\theta q}}+\dfrac{1}{\frac{r}{(1-\theta)q}}
\end{align*}
⟹ 1 = 1 = p θq + r ( 1 − θ ) q θq p 1 + ( 1 − θ ) q r 1
Thus, 1 p θ q \dfrac{1}{\frac{p}{\theta q}} θq p 1 and 1 r ( 1 − θ ) q \dfrac{1}{\frac{r}{(1-\theta)q}} ( 1 − θ ) q r 1 are conjugate exponents satisfying the Hölder’s inequality. Now, let each be as follows.
s = p q θ and s ′ = r ( 1 − θ ) q
s=\dfrac{p}{q\theta} \quad \text{and} \quad s^{\prime}=\dfrac{r}{(1-\theta)q}
s = qθ p and s ′ = ( 1 − θ ) q r
Hölder’s inequality
For any 1 p + 1 p ′ = 1 \dfrac{1}{p} + \dfrac{1}{p^{\prime}} = 1 p 1 + p ′ 1 = 1 being 1 ≤ p , p ′ < ∞ 1 \le p, p^{\prime} < \infty 1 ≤ p , p ′ < ∞ , if u ∈ L p ( Ω ) u \in L^p(\Omega) u ∈ L p ( Ω ) and v ∈ L p ′ ( Ω ) v\in L^{p^{\prime}}(\Omega) v ∈ L p ′ ( Ω ) , then
∫ Ω ∣ u ( x ) v ( x ) ∣ d x ≤ ∥ u ∥ p ∥ v ∥ p ′
\int_{\Omega} |u(x)v(x)| dx \le \| u \|_{p} \| v \|_{p^{\prime}}
∫ Ω ∣ u ( x ) v ( x ) ∣ d x ≤ ∥ u ∥ p ∥ v ∥ p ′
Case 1. r < ∞ r \lt \infty r < ∞
It can be verified that ∣ u ∣ θ q ∈ L s |u|^{\theta q} \in L^{s} ∣ u ∣ θq ∈ L s and ∣ u ∣ q ( 1 − θ ) ∈ L s ′ |u|^{q(1-\theta)} \in L^{s^{\prime}} ∣ u ∣ q ( 1 − θ ) ∈ L s ′ .
∥ u θ q ∥ s s = ∫ Ω ( ∣ u ( x ) ∣ q θ ) s d x = ∫ Ω ∣ u ( x ) ∣ p d x < ∞
\left\| u^{\theta q} \right\|_{s}^{s} = \int_{\Omega} \left( |u(x)|^{q\theta} \right)^{s} dx = \int_{\Omega} |u(x)|^p dx \lt \infty
u θq s s = ∫ Ω ( ∣ u ( x ) ∣ qθ ) s d x = ∫ Ω ∣ u ( x ) ∣ p d x < ∞
∥ u q ( 1 − θ ) ∥ s ′ s ′ = ∫ Ω ( ∣ u ( x ) ∣ q ( 1 − θ ) ) s ′ d x = ∫ Ω ∣ u ( x ) ∣ r d x < ∞
\left\| u^{q(1-\theta)} \right\|_{s^{\prime}}^{s^{\prime}}=\int_{\Omega} \left( |u(x)|^{q(1-\theta)} \right)^{s^{\prime}} dx=\int_{\Omega} |u(x)|^r dx \lt\infty
u q ( 1 − θ ) s ′ s ′ = ∫ Ω ( ∣ u ( x ) ∣ q ( 1 − θ ) ) s ′ d x = ∫ Ω ∣ u ( x ) ∣ r d x < ∞
Hence, we can use Hölder’s inequality. Calculating ∥ u ∥ q q \left\| u \right\|_{q}^{q} ∥ u ∥ q q yields
∥ u ∥ q q = ∫ Ω ∣ u ( x ) ∣ q d x = ∫ Ω ∣ u ( x ) ∣ θ q ∣ u ( x ) ∣ ( 1 − θ ) q d x ≤ ∥ u θ q ∥ s ∥ u ( 1 − θ ) q ∥ s ′ by Hoelder’s inequality = ( ∫ Ω ∣ u ( x ) ∣ θ q s d x ) 1 / s ( ∫ Ω ∣ u ( x ) ∣ ( 1 − θ ) q s ′ d x ) 1 / s ′ = ( ∫ Ω ∣ u ( x ) ∣ p d x ) 1 p q θ ( ∫ Ω ∣ u ( x ) ∣ r d x ) 1 r ( 1 − θ ) q = ∥ u ∥ p θ q ∥ u ∥ r ( 1 − θ ) q
\begin{align*}
\left\| u \right\|_{q}^{q} =&\ \int_{\Omega} |u(x)|^{q} dx
\\ =&\ \int_{\Omega} |u(x)|^{\theta q} |u(x)|^{(1-\theta)q} dx
\\ \le& \left\| u^{\theta q} \right\|_{s} \left\| u^{(1-\theta)q} \right\|_{s^{\prime}} \quad \text{by Hoelder’s inequality}
\\ =&\ \left( \int_{\Omega} |u(x)|^{\theta q s } dx \right)^{1/s} \left( \int_{\Omega} |u(x)|^{(1-\theta) q s^{\prime} } dx \right)^{1/s^{\prime}}
\\ =&\ \left( \int_{\Omega} |u(x)|^{p} dx \right)^{\frac{1}{p}q\theta} \left( \int_{\Omega} |u(x)|^{r} dx \right)^{\frac{1}{r}(1-\theta)q}
\\ =&\ \| u \|_{p}^{\theta q} \left\| u \right\|_{r}^{(1-\theta)q}
\end{align*}
∥ u ∥ q q = = ≤ = = = ∫ Ω ∣ u ( x ) ∣ q d x ∫ Ω ∣ u ( x ) ∣ θq ∣ u ( x ) ∣ ( 1 − θ ) q d x u θq s u ( 1 − θ ) q s ′ by Hoelder’s inequality ( ∫ Ω ∣ u ( x ) ∣ θq s d x ) 1/ s ( ∫ Ω ∣ u ( x ) ∣ ( 1 − θ ) q s ′ d x ) 1/ s ′ ( ∫ Ω ∣ u ( x ) ∣ p d x ) p 1 qθ ( ∫ Ω ∣ u ( x ) ∣ r d x ) r 1 ( 1 − θ ) q ∥ u ∥ p θq ∥ u ∥ r ( 1 − θ ) q
By multiplying the exponent of both sides by 1 q \dfrac{1}{q} q 1 , we get
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ r 1 − θ
\left\| u \right\|_{q} \le \left\| u \right\|_{p}^{\theta} \left\| u \right\|_{r}^{1-\theta}
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ r 1 − θ
Case 2. r = ∞ r = \infty r = ∞
The condition of the assumption is such that 1 q = θ p \dfrac{1}{q} = \dfrac{\theta}{p} q 1 = p θ . Similarly, if we let s = p θ q = 1 s=\frac{p}{\theta q}=1 s = θq p = 1 having s ′ = ∞ s^{\prime}=\infty s ′ = ∞ and satisfying 1 = 1 s + 1 s ′ 1=\frac{1}{s}+\frac{1}{s^{\prime}} 1 = s 1 + s ′ 1 with ∣ u ∣ θ q ∈ L s = L 1 |u|^{\theta q} \in L^s=L^1 ∣ u ∣ θq ∈ L s = L 1 and ∣ u ∣ ( 1 − θ ) q ∈ L s ′ = L ∞ |u|^{(1-\theta)q} \in L^{s^{\prime}}=L^{\infty} ∣ u ∣ ( 1 − θ ) q ∈ L s ′ = L ∞ .
∥ u θ q ∥ 1 = ∫ Ω ∣ u ( x ) ∣ q θ d x = ∫ Ω ∣ u ( x ) ∣ p d x < ∞
\left\| u^{\theta q} \right\|_{1}=\int_{\Omega} |u(x)|^{q\theta} dx=\int_{\Omega} |u(x)|^p dx \lt\infty
u θq 1 = ∫ Ω ∣ u ( x ) ∣ qθ d x = ∫ Ω ∣ u ( x ) ∣ p d x < ∞
If u ∈ L ∞ u\in L^{\infty} u ∈ L ∞ , it is easily understood by the definition of ∣ ⋅ ∣ ∞ | \cdot|_\infty ∣ ⋅ ∣ ∞ that for any positive number k k k , ∣ u ∣ k ∈ L ∞ |u|^k \in L^{\infty} ∣ u ∣ k ∈ L ∞ holds. Since ( 1 − θ ) > 0 (1-\theta)\gt 0 ( 1 − θ ) > 0 , thus ∣ u ∣ ( 1 − θ ) q ∈ L ∞ |u|^{(1-\theta)q} \in L^{\infty} ∣ u ∣ ( 1 − θ ) q ∈ L ∞ . Therefore, using Hölder’s inequality,
∥ u ∥ q q = ∫ Ω ∣ u ∣ q d x = ∫ Ω ∣ u ∣ θ q ∣ u ∣ ( 1 − θ ) q d x ≤ ∥ u θ q ∥ 1 ∥ u ( 1 − θ ) q ∥ ∞ = ( ∫ Ω ∣ u ∣ θ q d x ) 1 ∥ u ( 1 − θ ) q ∥ ∞ = ( ∫ Ω ∣ u ∣ p d x ) 1 p q θ ∥ u ( 1 − θ ) q ∥ ∞ = ∥ u ∥ p θ q ∥ u ( 1 − θ ) q ∥ ∞ = ∥ u ∥ p θ q ∥ u ∥ ∞ ( 1 − θ ) q
\begin{align*}
\left\| u \right\|_{q}^{q} =&\ \int_{\Omega} |u|^q dx
\\ =&\ \int_{\Omega} |u|^{\theta q} |u|^{(1-\theta)q}dx
\\ \le& \left\| u^{\theta q} \right\|_{1} \left\| u^{(1-\theta)q} \right\|_{\infty}
\\ =&\ \left( \int_{\Omega} |u|^{\theta q} dx \right)^{1} \left\| u^{(1-\theta)q} \right\|_{\infty}
\\ =&\ \left( \int_{\Omega} |u|^{p} dx \right)^{\frac{1}{p}q\theta} \left\| u^{(1-\theta)q} \right\|_{\infty}
\\ =&\ \| u \|_{p}^{\theta q} \left\| u^{(1-\theta)q} \right\|_{\infty}
\\ =&\ \| u \|_{p}^{\theta q}\ \left\| u \right\|_{\infty}^{(1-\theta)q}
\end{align*}
∥ u ∥ q q = = ≤ = = = = ∫ Ω ∣ u ∣ q d x ∫ Ω ∣ u ∣ θq ∣ u ∣ ( 1 − θ ) q d x u θq 1 u ( 1 − θ ) q ∞ ( ∫ Ω ∣ u ∣ θq d x ) 1 u ( 1 − θ ) q ∞ ( ∫ Ω ∣ u ∣ p d x ) p 1 qθ u ( 1 − θ ) q ∞ ∥ u ∥ p θq u ( 1 − θ ) q ∞ ∥ u ∥ p θq ∥ u ∥ ∞ ( 1 − θ ) q
The third line is established using Hölder’s inequality. The last line is established by the property of ∥ ⋅ ∥ ∞ \left\| \cdot \right\| _{\infty} ∥ ⋅ ∥ ∞ . Hence,
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ ∞ ( 1 − θ ) = ∥ u ∥ p θ ∣ u ∣ r ( 1 − θ )
\left\| u \right\|_{q} \le \| u \|_{p}^{\theta } \left\| u \right\|_{\infty}^{(1-\theta)}= \| u \|_{p}^{\theta }\ |u|_{r}^{(1-\theta)}
∥ u ∥ q ≤ ∥ u ∥ p θ ∥ u ∥ ∞ ( 1 − θ ) = ∥ u ∥ p θ ∣ u ∣ r ( 1 − θ )
■