Brezis Functional Analysis: Mastering W^{1,p}(\mathbb{R}) Spaces

by Andrew McMorgan 65 views

Hey guys! Today, we're diving deep into a classic problem from Brezis' Functional Analysis, specifically Exercise 8.1. We're going to dissect a function and figure out for which values of p it belongs to the Sobolev space W1,p(R)W^{1,p}(\mathbb{R}). This is a super important concept in functional analysis, dealing with functions that not only have certain integrability properties but also possess derivatives in certain spaces. So, grab your coffee, get comfy, and let's break down this challenging exercise. We'll be looking at the function u(x)=1(1+x2)Ξ±21ln⁑(2+x2)u(x) = \frac{1}{(1+x^2)^{\frac{\alpha}{2}}} \frac{1}{\ln(2+ x^2)} with 0<Ξ±<10 < \alpha < 1, and our mission is to determine for which p∈[1,∞)p \in [1, \infty) this function uu is in W1,p(R)W^{1,p}(\mathbb{R}). Remember, for uu to be in W1,p(R)W^{1,p}(\mathbb{R}), both uu itself and its weak derivative (which, for this well-behaved function, will be its classical derivative) must be in Lp(R)L^p(\mathbb{R}). This means we need to check the integrability of uu and its derivative uβ€²u'. Let's get started by understanding what W1,p(R)W^{1,p}(\mathbb{R}) actually means. The Sobolev space W1,p(R)W^{1,p}(\mathbb{R}) consists of all functions uinLp(R)u in L^p(\mathbb{R}) whose first weak derivative uβ€²u' also belongs to Lp(R)L^p(\mathbb{R}). The norm in this space is typically defined as ∣∣u∣∣W1,p=(∣∣u∣∣Lpp+∣∣uβ€²βˆ£βˆ£Lpp)1p||u||_{W^{1,p}} = (||u||_{L^p}^p + ||u'||_{L^p}^p)^{\frac{1}{p}}. For our specific function u(x)u(x), which is smooth and decays at infinity, its weak derivative will coincide with its classical derivative. So, the task boils down to calculating uβ€²(x)u'(x) and then investigating whether both uu and uβ€²u' are pp-integrable over the entire real line. This involves analyzing the behavior of u(x)u(x) and uβ€²(x)u'(x) as xβ†’Β±βˆžx \to \pm \infty and near any potential singularities. In our case, the function is well-defined and smooth everywhere, so we're mainly concerned with its decay rate. This exercise is a fantastic way to solidify your understanding of function spaces and how to analyze functions within them. Let's roll up our sleeves and tackle the derivative calculation, which is the first crucial step in this investigation. It's going to involve the chain rule and the quotient rule, so keep those differentiation techniques sharp, guys!

Calculating the Derivative uβ€²(x)u'(x)

Alright guys, the first major step in checking if our function u(x)u(x) belongs to W1,p(R)W^{1,p}(\mathbb{R}) is to compute its derivative, uβ€²(x)u'(x). For u(x)=1(1+x2)Ξ±21ln⁑(2+x2)u(x) = \frac{1}{(1+x^2)^{\frac{\alpha}{2}}} \frac{1}{\ln(2+ x^2)}, we'll need to use the product rule and the chain rule. Let's rewrite u(x)u(x) slightly to make the differentiation process a bit more straightforward: u(x)=(1+x2)βˆ’Ξ±2(ln⁑(2+x2))βˆ’1u(x) = (1+x^2)^{-\frac{\alpha}{2}} (\ln(2+x^2))^{-1}. Now, applying the product rule (fg)β€²=fβ€²g+fgβ€²(fg)' = f'g + fg':

Let f(x)=(1+x2)βˆ’Ξ±2f(x) = (1+x^2)^{-\frac{\alpha}{2}} and g(x)=(ln⁑(2+x2))βˆ’1g(x) = (\ln(2+x^2))^{-1}.

First, let's find fβ€²(x)f'(x): Using the chain rule, ddx(an)=nanβˆ’1β‹…aβ€²\frac{d}{dx}(a^n) = na^{n-1} \cdot a', where a=1+x2a = 1+x^2 and n=βˆ’Ξ±2n = -\frac{\alpha}{2}. So, aβ€²=2xa' = 2x. fβ€²(x)=βˆ’Ξ±2(1+x2)βˆ’Ξ±2βˆ’1β‹…(2x)=βˆ’Ξ±x(1+x2)βˆ’Ξ±2βˆ’1f'(x) = -\frac{\alpha}{2}(1+x^2)^{-\frac{\alpha}{2}-1} \cdot (2x) = -\alpha x (1+x^2)^{-\frac{\alpha}{2}-1}.

Next, let's find gβ€²(x)g'(x): Using the chain rule and the rule for the derivative of ln⁑(v)\ln(v), which is vβ€²v\frac{v'}{v}. Here, v=2+x2v = 2+x^2, so vβ€²=2xv' = 2x. The derivative of yβˆ’1y^{-1} is βˆ’yβˆ’2-y^{-2}. So, $g'(x) = - (\ln(2+x2)){-2} \cdot \frac{1}{\ln(2+x^2)}' $. We need the derivative of (ln⁑(2+x2))βˆ’1(\ln(2+x^2))^{-1}. Let w=ln⁑(2+x2)w = \ln(2+x^2). Then g(x)=wβˆ’1g(x) = w^{-1}, and gβ€²(x)=βˆ’wβˆ’2β‹…wβ€²g'(x) = -w^{-2} \cdot w'. The derivative of w=ln⁑(2+x2)w = \ln(2+x^2) is 2x2+x2\frac{2x}{2+x^2}. So, gβ€²(x)=βˆ’(ln⁑(2+x2))βˆ’2β‹…2x2+x2=βˆ’2x(2+x2)(ln⁑(2+x2))2g'(x) = -(\ln(2+x^2))^{-2} \cdot \frac{2x}{2+x^2} = -\frac{2x}{(2+x^2)(\ln(2+x^2))^2}.

Now, let's put it all together using the product rule: uβ€²(x)=fβ€²(x)g(x)+f(x)gβ€²(x)u'(x) = f'(x)g(x) + f(x)g'(x).

uβ€²(x)=(βˆ’Ξ±x(1+x2)βˆ’Ξ±2βˆ’1)((ln⁑(2+x2))βˆ’1)+((1+x2)βˆ’Ξ±2)(βˆ’2x(2+x2)(ln⁑(2+x2))2)u'(x) = \left( -\alpha x (1+x^2)^{-\frac{\alpha}{2}-1} \right) \left( (\ln(2+x^2))^{-1} \right) + \left( (1+x^2)^{-\frac{\alpha}{2}} \right) \left( -\frac{2x}{(2+x^2)(\ln(2+x^2))^2} \right)

uβ€²(x)=βˆ’Ξ±x(1+x2)Ξ±2+1ln⁑(2+x2)βˆ’2x(1+x2)Ξ±2(2+x2)(ln⁑(2+x2))2u'(x) = -\frac{\alpha x}{(1+x^2)^{\frac{\alpha}{2}+1} \ln(2+x^2)} - \frac{2x}{(1+x^2)^{\frac{\alpha}{2}}(2+x^2)(\ln(2+x^2))^2}

We can factor out some common terms to simplify this expression. Let's factor out βˆ’x(1+x2)βˆ’Ξ±2βˆ’1(ln⁑(2+x2))βˆ’2-x (1+x^2)^{-\frac{\alpha}{2}-1} (\ln(2+x^2))^{-2}:

uβ€²(x)=βˆ’x(1+x2)βˆ’Ξ±2βˆ’1(ln⁑(2+x2))βˆ’2[Ξ±(ln⁑(2+x2))+2(1+x2)(2+x2)]u'(x) = -x (1+x^2)^{-\frac{\alpha}{2}-1} (\ln(2+x^2))^{-2} \left[ \alpha (\ln(2+x^2)) + \frac{2(1+x^2)}{(2+x^2)} \right]

This looks a bit messy, but that's okay. The key is to understand the asymptotic behavior of this derivative as xβ†’Β±βˆžx \to \pm \infty. This derivative calculation is a critical step, and getting it right ensures we're on the right track for analyzing the integrability of uβ€²u'. Don't worry if it looks complicated; we'll simplify our analysis by focusing on the dominant terms for large ∣x∣|x|. The next stage is to investigate the integrability of both u(x)u(x) and uβ€²(x)u'(x) over R\mathbb{R}.

Analyzing the Integrability of u(x)u(x)

Now that we've got the derivative uβ€²(x)u'(x) all worked out, let's shift our focus to the integrability of the original function u(x)u(x) and its derivative. For uu to be in W1,p(R)W^{1,p}(\mathbb{R}), both uinLp(R)u in L^p(\mathbb{R}) and uβ€²inLp(R)u' in L^p(\mathbb{R}). Let's start with u(x)=1(1+x2)Ξ±21ln⁑(2+x2)u(x) = \frac{1}{(1+x^2)^{\frac{\alpha}{2}}} \frac{1}{\ln(2+ x^2)}, where 0<Ξ±<10 < \alpha < 1. We need to check if βˆ«βˆ’βˆžβˆžβˆ£u(x)∣pdx<∞\int_{-\infty}^{\infty} |u(x)|^p dx < \infty for different values of p∈[1,∞)p \in [1, \infty).

Let's consider the behavior of u(x)u(x) as ∣xβˆ£β†’βˆž|x| \to \infty. As ∣x∣|x| becomes large, 1+x2β‰ˆx21+x^2 \approx x^2 and 2+x2β‰ˆx22+x^2 \approx x^2. So, for large ∣x∣|x|, u(x)u(x) behaves like:

u(x)∼1(x2)α21ln⁑(x2)=1∣x∣α12ln⁑∣x∣u(x) \sim \frac{1}{(x^2)^{\frac{\alpha}{2}}} \frac{1}{\ln(x^2)} = \frac{1}{|x|^{\alpha}} \frac{1}{2\ln|x|}.

Now, let's consider the integral ∫∞∣u(x)∣pdx\int^{\infty} |u(x)|^p dx. For large xx, this integral behaves like:

∫∞∣1xα12ln⁑x∣pdx=∫∞1∣x∣pα1(2ln⁑x)pdx\int^{\infty} \left| \frac{1}{x^{\alpha}} \frac{1}{2\ln x} \right|^p dx = \int^{\infty} \frac{1}{|x|^{p\alpha}} \frac{1}{(2\ln x)^p} dx.

For this integral to converge, we need the power of xx in the denominator to be greater than 1. That is, we need pΞ±>1p\alpha > 1. This condition gives us a lower bound on pp for the integrability of u(x)u(x). If pα≀1p\alpha \le 1, the integral will diverge.

Let's think about the condition pΞ±>1p\alpha > 1. Since 0<Ξ±<10 < \alpha < 1, if pΞ±>1p\alpha > 1, it means that p>1Ξ±p > \frac{1}{\alpha}. Since Ξ±<1\alpha < 1, 1Ξ±>1\frac{1}{\alpha} > 1. So, this condition automatically implies p>1p > 1. If p=1p=1, the condition becomes Ξ±>1\alpha > 1, which is not true. So for p=1p=1, we need to be careful.

If pΞ±>1p\alpha > 1, the term ∣xβˆ£βˆ’pΞ±|x|^{-p\alpha} dominates the slower decaying term (ln⁑x)βˆ’p(\ln x)^{-p}. The integral ∫∞∣xβˆ£βˆ’pΞ±dx\int^{\infty} |x|^{-p\alpha} dx converges if and only if pΞ±>1p\alpha > 1. Thus, for u(x)u(x) to be in Lp(R)L^p(\mathbb{R}), we require pΞ±>1p\alpha > 1, which implies p>1Ξ±p > \frac{1}{\alpha}.

What happens if pα≀1p\alpha \le 1? If pΞ±=1p\alpha = 1, then u(x)∼1∣x∣12ln⁑∣x∣u(x) \sim \frac{1}{|x|} \frac{1}{2\ln|x|}. The integral ∫∞1xln⁑xdx\int^{\infty} \frac{1}{x \ln x} dx diverges (by substitution y=ln⁑xy=\ln x, dy=1xdxdy = \frac{1}{x} dx, so ∫1ydy=ln⁑∣y∣=ln⁑∣ln⁑x∣\int \frac{1}{y} dy = \ln|y| = \ln|\ln x|). So, if pΞ±=1p\alpha = 1, u(x)u(x) is not in Lp(R)L^p(\mathbb{R}).

If pΞ±<1p\alpha < 1, then ∣xβˆ£βˆ’pΞ±|x|^{-p\alpha} grows slower than ∣xβˆ£βˆ’1|x|^{-1}, and the integral ∫∞∣xβˆ£βˆ’pΞ±dx\int^{\infty} |x|^{-p\alpha} dx diverges even faster.

Therefore, for u(x)u(x) to be in Lp(R)L^p(\mathbb{R}), we must have pΞ±>1p\alpha > 1, which means p>1Ξ±p > \frac{1}{\alpha}. Since 0<Ξ±<10 < \alpha < 1, we know 1Ξ±>1\frac{1}{\alpha} > 1. So, this condition ensures p>1p > 1.

This analysis tells us that for u(x)u(x) to be pp-integrable, we need p>1Ξ±p > \frac{1}{\alpha}. The behavior near x=0x=0 is not an issue because u(x)u(x) is bounded and well-behaved there. The crucial part is the decay at infinity.

Analyzing the Integrability of uβ€²(x)u'(x)

Now, let's get serious about the derivative uβ€²(x)u'(x). Remember, uβ€²(x)=βˆ’Ξ±x(1+x2)Ξ±2+1ln⁑(2+x2)βˆ’2x(1+x2)Ξ±2(2+x2)(ln⁑(2+x2))2u'(x) = -\frac{\alpha x}{(1+x^2)^{\frac{\alpha}{2}+1} \ln(2+x^2)} - \frac{2x}{(1+x^2)^{\frac{\alpha}{2}}(2+x^2)(\ln(2+x^2))^2}. We need to check if βˆ«βˆ’βˆžβˆžβˆ£uβ€²(x)∣pdx<∞\int_{-\infty}^{\infty} |u'(x)|^p dx < \infty. Again, we focus on the behavior as ∣xβˆ£β†’βˆž|x| \to \infty.

As ∣xβˆ£β†’βˆž|x| \to \infty, 1+x2β‰ˆx21+x^2 \approx x^2 and 2+x2β‰ˆx22+x^2 \approx x^2. Let's analyze the dominant terms in uβ€²(x)u'(x).

The first term behaves like: βˆ’Ξ±x(x2)Ξ±2+1ln⁑(x2)=βˆ’Ξ±xxΞ±+2(2ln⁑∣x∣)=βˆ’Ξ±2xΞ±+1ln⁑∣x∣- \frac{\alpha x}{(x^2)^{\frac{\alpha}{2}+1} \ln(x^2)} = - \frac{\alpha x}{x^{\alpha+2} (2\ln|x|)} = - \frac{\alpha}{2x^{\alpha+1} \ln|x|}.

The second term behaves like: βˆ’2x(x2)Ξ±2(x2)(ln⁑(x2))2=βˆ’2xxΞ±x2(2ln⁑∣x∣)2=βˆ’2xxΞ±+2(4(ln⁑∣x∣)2)=βˆ’12xΞ±+1(ln⁑∣x∣)2- \frac{2x}{(x^2)^{\frac{\alpha}{2}}(x^2)(\ln(x^2))^2} = - \frac{2x}{x^{\alpha} x^2 (2\ln|x|)^2} = - \frac{2x}{x^{\alpha+2} (4(\ln|x|)^2)} = - \frac{1}{2x^{\alpha+1} (\ln|x|)^2}.

So, for large ∣x∣|x|, uβ€²(x)u'(x) behaves like:

uβ€²(x)βˆΌβˆ’Ξ±2xΞ±+1ln⁑∣xβˆ£βˆ’12xΞ±+1(ln⁑∣x∣)2u'(x) \sim -\frac{\alpha}{2x^{\alpha+1} \ln|x|} - \frac{1}{2x^{\alpha+1} (\ln|x|)^2}.

The term with the higher power in the logarithm dominates the decay. Since (ln⁑∣x∣)2(\ln|x|)^2 grows faster than ln⁑∣x∣\ln|x|, the second term decays faster. Therefore, the dominant term is βˆ’Ξ±2xΞ±+1ln⁑∣x∣-\frac{\alpha}{2x^{\alpha+1} \ln|x|}.

So, uβ€²(x)βˆΌβˆ’Ξ±21xΞ±+1ln⁑∣x∣u'(x) \sim -\frac{\alpha}{2} \frac{1}{x^{\alpha+1} \ln|x|} for large ∣x∣|x|.

Now we need to check the integrability of ∣uβ€²(x)∣p|u'(x)|^p. For large xx, this behaves like:

βˆ«βˆžβˆ£βˆ’Ξ±21xΞ±+1ln⁑x∣pdx=∫∞(Ξ±2)p1∣x∣p(Ξ±+1)(ln⁑x)pdx\int^{\infty} \left| -\frac{\alpha}{2} \frac{1}{x^{\alpha+1} \ln x} \right|^p dx = \int^{\infty} \left( \frac{\alpha}{2} \right)^p \frac{1}{|x|^{p(\alpha+1)} (\ln x)^p} dx.

For this integral to converge, we need the power of xx in the denominator to be greater than 1. That is, p(Ξ±+1)>1p(\alpha+1) > 1.

This condition p(α+1)>1p(\alpha+1) > 1 implies p>1α+1p > \frac{1}{\alpha+1}. Since 0<α<10 < \alpha < 1, we have 1<α+1<21 < \alpha+1 < 2. Thus, 12<1α+1<1\frac{1}{2} < \frac{1}{\alpha+1} < 1. This condition alone doesn't guarantee p>1p > 1, which is necessary for the LpL^p norm. However, we are considering p∈[1,∞)p \in [1, \infty).

Let's analyze the condition p(Ξ±+1)>1p(\alpha+1) > 1 more closely. If p(Ξ±+1)>1p(\alpha+1) > 1, then ∫∞∣xβˆ£βˆ’p(Ξ±+1)dx\int^{\infty} |x|^{-p(\alpha+1)} dx converges. The logarithmic term (ln⁑x)p(\ln x)^p decays slower than any power of xx, so its presence doesn't prevent convergence if the power of xx is greater than 1.

So, the condition for uβ€²(x)u'(x) to be in Lp(R)L^p(\mathbb{R}) boils down to p(Ξ±+1)>1p(\alpha+1) > 1, which simplifies to p>1Ξ±+1p > \frac{1}{\alpha+1}.

Let's check the edge case p=1p=1. If p=1p=1, the condition becomes Ξ±+1>1\alpha+1 > 1, which is Ξ±>0\alpha > 0. This is true since 0<Ξ±<10 < \alpha < 1. So, uβ€²(x)u'(x) is integrable for p=1p=1 as long as Ξ±>0\alpha > 0. The integral ∫∞1xΞ±+1ln⁑xdx\int^{\infty} \frac{1}{x^{\alpha+1} \ln x} dx converges because Ξ±+1>1\alpha+1 > 1.

Therefore, for uβ€²(x)u'(x) to be in Lp(R)L^p(\mathbb{R}), we require p>1Ξ±+1p > \frac{1}{\alpha+1}.

Bringing It All Together: The Condition for u∈W1,p(R)u \in W^{1,p}(\mathbb{R})

We've analyzed the integrability of both u(x)u(x) and uβ€²(x)u'(x). For uu to be in the Sobolev space W1,p(R)W^{1,p}(\mathbb{R}), both conditions must be satisfied:

  1. For u∈Lp(R)u \in L^p(\mathbb{R}), we need p>1αp > \frac{1}{\alpha}.
  2. For uβ€²βˆˆLp(R)u' \in L^p(\mathbb{R}), we need p>1Ξ±+1p > \frac{1}{\alpha+1}.

We need to find the values of p∈[1,∞)p \in [1, \infty) that satisfy both of these inequalities.

Let's compare the two lower bounds: 1Ξ±\frac{1}{\alpha} and 1Ξ±+1\frac{1}{\alpha+1}. Since 0<Ξ±<10 < \alpha < 1, we have Ξ±+1>Ξ±\alpha+1 > \alpha. Taking reciprocals reverses the inequality, so 1Ξ±+1<1Ξ±\frac{1}{\alpha+1} < \frac{1}{\alpha}.

This means that the condition p>1Ξ±p > \frac{1}{\alpha} is stricter than p>1Ξ±+1p > \frac{1}{\alpha+1}. If pp satisfies p>1Ξ±p > \frac{1}{\alpha}, it automatically satisfies p>1Ξ±+1p > \frac{1}{\alpha+1} because 1Ξ±>1Ξ±+1\frac{1}{\alpha} > \frac{1}{\alpha+1}.

Therefore, the combined condition for uinW1,p(R)u in W^{1,p}(\mathbb{R}) is simply p>1Ξ±p > \frac{1}{\alpha}.

Since 0<Ξ±<10 < \alpha < 1, we have 1Ξ±>1\frac{1}{\alpha} > 1. So, the condition p>1Ξ±p > \frac{1}{\alpha} means that pp must be strictly greater than some value greater than 1.

Thus, uinW1,p(R)u in W^{1,p}(\mathbb{R}) for all p∈(1α,∞)p \in \left( \frac{1}{\alpha}, \infty \right).

Remember that the problem asked for p∈[1,∞)p \in [1, \infty). Our result p>1αp > \frac{1}{\alpha} falls within this range because 1α>1\frac{1}{\alpha} > 1 (since α<1\alpha < 1).

So, the final answer is that uinW1,p(R)u in W^{1,p}(\mathbb{R}) for all pp such that p>1Ξ±p > \frac{1}{\alpha}. This is a really cool result, guys! It shows how the parameters in the function, like Ξ±\alpha, dictate which function spaces it belongs to. It’s all about the decay rates of the function and its derivatives. Keep practicing these types of problems, and you'll become a pro at analyzing functions in Sobolev spaces!

Conclusion: Mastering Sobolev Spaces

We've successfully navigated through Brezis' Functional Analysis Exercise 8.1, determining the range of p for which our function u(x)=1(1+x2)Ξ±21ln⁑(2+x2)u(x) = \frac{1}{(1+x^2)^{\frac{\alpha}{2}}} \frac{1}{\ln(2+ x^2)} with 0<Ξ±<10 < \alpha < 1 belongs to the Sobolev space W1,p(R)W^{1,p}(\mathbb{R}). The key takeaway is that for uu to be in W1,p(R)W^{1,p}(\mathbb{R}), both the function itself and its derivative must be pp-integrable over R\mathbb{R}. Through careful asymptotic analysis of u(x)u(x) and uβ€²(x)u'(x) as ∣xβˆ£β†’βˆž|x| \to \infty, we found that u∈Lp(R)u \in L^p(\mathbb{R}) requires p>1Ξ±p > \frac{1}{\alpha}, and uβ€²βˆˆLp(R)u' \in L^p(\mathbb{R}) requires p>1Ξ±+1p > \frac{1}{\alpha+1}. Since 1Ξ±>1Ξ±+1\frac{1}{\alpha} > \frac{1}{\alpha+1} for 0<Ξ±<10 < \alpha < 1, the stricter condition p>1Ξ±p > \frac{1}{\alpha} dictates the range. Therefore, uinW1,p(R)u in W^{1,p}(\mathbb{R}) for all p∈(1Ξ±,∞)p \in \left( \frac{1}{\alpha}, \infty \right).

This exercise highlights the fundamental importance of understanding the decay rates of functions and their derivatives when working with function spaces like Sobolev spaces. The parameter Ξ±\alpha plays a crucial role in defining the integrability properties of uu and uβ€²u'. This kind of analysis is fundamental in various areas of mathematics and physics, including the study of partial differential equations. Keep practicing these detailed analyses, guys. The more you work through these examples, the more intuitive the concepts of function spaces and their properties will become. Functional analysis might seem daunting at first, but with persistence and a good understanding of the core principles, you can master these challenging topics. Keep up the great work, and happy analyzing!