3

Suppose that $f: \mathbb{R} \rightarrow \mathbb{R}$ is a continuous function. Write down the negation of the statement $$\exists \alpha>0 \quad \forall s>0 \quad \exists x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha} d t\right|<s.\tag1$$ Prove whether statement $(1)$ or its negation is true.

Symbolically, the negation of statement $(1)$ is $$ \forall \alpha>0 \quad \exists s>0 \quad \forall x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha} d t\right|\geqslant s . $$

I feel like the negation is easy to disprove as you can just have $f(t)=0$. However, I believe that disproving the negation is not enough to prove the original statement in this case, as it is only showing the original statement to be true for the specific case $f=0,$ not for all $f.$ Our lecturer has confirmed this to be true.

I am unsure of how to proceed with the general function for a proof of the original statement.

Lucas
  • 33
  • Think through the quantifiers. If you have a particular continuous $f$ for which the negation is false, then you are done. I recommend that you write an answer. – Ted Shifrin Mar 21 '23 at 01:29
  • Is it enough just to show that the negation is false? I feel like that is too easy as you can just choose $f=0$ which can clearly not be greater than any $s>0$. – Lucas Mar 21 '23 at 01:42
  • The "suppose $f$" of the statement acts as a "$\forall f$", which gives a "$\exists f$" in the negation. Thus to disprove the negation, you can't fix some $f$. Your negation of $a < b$ is also wrong. – caduk Mar 23 '23 at 13:18
  • @ryang I know that it is not enough to disprove the negation. I am just really unsure of how to even begin proving the original statement. – Lucas Mar 23 '23 at 22:18
  • @caduk I am not sure what you mean by "Your negation of $a<b$ is also wrong? – Lucas Mar 23 '23 at 22:20
  • The negation of $a < b$ is not $a > b$ but $a \geq a$. – caduk Mar 23 '23 at 22:34
  • @ryang I believe that disproving the negation is not enough to prove the original statement in this case as it is only showing the original statement to be true for the specific case $f=0$, not for all $f$. Our lecturer has confirmed this to be true also. I have also been provided with the hint: "once you pick your $x$ then your function is continuous on a closed bounded interval". I still don't know how this helps though! – Lucas Mar 24 '23 at 06:36
  • Assume $f: \mathbb{R} \rightarrow \mathbb{R}$ is a continuous function. Consider the statement $$ \exists \alpha>0 \quad \forall s>0 \quad \exists x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha} d t\right|<s $$ $$ \text{(a) Write the negation of the above statement.} $$ (b) Which is true, the statement or its negation? Prove your answer.

    This is a straight cut and paste (using MathPix snipping tool) from the homework set.

    – Lucas Mar 24 '23 at 09:22
  • @ryang Yes, I think I have made sense of it now and have some understanding that is enough to go on. Thank you! – Lucas Mar 25 '23 at 03:32

1 Answers1

2

Suppose that $f: \mathbb{R} \rightarrow \mathbb{R}$ is a continuous function. Write down the negation of the statement $$\exists \alpha>0 \quad \forall s>0 \quad \exists x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha} d t\right|<s.\tag1$$ Prove whether statement $(1)$ or its negation is true.

Symbolically, the negation of statement $(1)$ is $$ \forall \alpha>0 \quad \exists s>0 \quad \forall x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha} d t\right|\geqslant s.\tag{1n}$$

I feel like the negation is easy to disprove as you can just have $f(t)=0$. However, I believe that disproving the negation is not enough to prove the original statement in this case, as it is only showing the original statement to be true for the specific case $f=0,$ not for all $f.$ Our lecturer has confirmed this to be true.

Please understand that $(1)$ and its negation $(1\text n)$ are not statements, as they contain the free variable $f;$ here, neither happens to have a definite truth value. Let $C$ denote the set of all continuous functions from $\mathbb R$ to $\mathbb R.$ Now, these are statements: $$\forall f\in C\quad \exists \alpha>0 \quad \forall s>0 \quad \exists x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha}\,\mathrm d t\right|<s.\tag2$$ $$\forall f\in C\quad\forall \alpha>0 \quad \exists s>0 \quad \forall x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha}\,\mathrm d t\right|\geqslant s.\tag3$$ $$\exists f\in C\quad\forall \alpha>0 \quad \exists s>0 \quad \forall x>1 \quad\left|\int_1^x \frac{f(t)}{(t-1)^\alpha}\,\mathrm d t\right|\geqslant s.\tag{2n}$$

Exhibiting $f\equiv0$ disproves statement $(3);$ however, it is $(2\text n)$ rather than $(3)$ that negates $(2).$

I believe that disproving the negation is not enough to prove the original statement in this case

No: proving statement P is in fact logically equivalent to disproving its negation.

Does your lecturer actually want you to choose between proving $(2)$ and $(2\text n) ?$ This is not actually how they have written their exercise.

In any case, we can prove statement $(2)$ like this. Put $\alpha=1$ and observe that the integrand is continuous on $(1,\infty);$ so, by the FTC, for each $b$ and $x$ such that $x>b>1,$ for some antiderivative $F,$ $\left|\int_b^x \frac{f(t)}{(t-1)^\alpha}\,\mathrm d t\right|$ equals $|F(x)-F(b)|;$ we want to show that $\displaystyle\lim_{x\to b^+}\lim_{b\to1^+}|F(x)-F(b)|=0$ that is, $\displaystyle\lim_{x\to b^+}\lim_{b\to1^+}F(x)-F(b)=0;$ this is indeed true, by definition, since the antiderivative $F$ is continuous on $(1,\infty).$

ryang
  • 38,879
  • 14
  • 81
  • 179
  • Regarding (2n): Suppose there existed such an $f\in C$, then it is enough to consider $\alpha>0$ s.t. the integral converges about $t=1$ (otherwise the lhs is clearly $\geq s$). In this case you can choose $x$ close enough to $1$ s.t. the inequality is violated for every fixed $s>0$. – Diger Mar 24 '23 at 19:43
  • Your proof doesn't work for $\alpha=1$ I think, because the integral does not converge. $F(1)$ will not be defined. Though you can just take $\alpha=1/2$ instead. – Diger Mar 24 '23 at 21:26
  • @Diger Thanks for pointing that out (how does taking $\alpha=\frac12$ fix the issue though?); I've just fixed the proof, I think. – ryang Mar 25 '23 at 00:21
  • @ryang I wonder though, if it is enough to show that $\lim_{x\to b^+}\lim_{b\to 1^+}F(x)-F(b)=0$ since we can also have $s$ arbitrarily close to zero. Couldn't we just set $s$ to be some function which converges to zero faster than $F(x)-F(b)$? Though I must confess that most of this is beyond me. – Lucas Mar 25 '23 at 05:17
  • 1
    @Lucas 1. You're forgetting that the choice of $x$ depends on the choice of $s$ (we know this from the order of their quantifiers). $\quad$ 2. $(2)$ and $(2\text n)$ are not propositions either, but open formulae. $\quad$ 3. More charitably, the exercise either actually means for you to choose between $(2)$ and $(3)$ (in which case it is wrong that their logical forms return opposite truth values), or it actually means for you to choose between $(2)$ and $(2\text n)$ (in which case the bit about $f$ needs to shift into the original statement). – ryang Mar 25 '23 at 07:03
  • 1
    @ryang I have typed up an answer that is similar to yours with the limits, so I will see how that goes and whether the lecturer accepts it. Thank you again, I have learnt more through trying, failing and discussing this problem here than I perhaps would have otherwise. – Lucas Mar 25 '23 at 07:13
  • @Ryang: Well, for $\alpha<1$ - for instance $\alpha=1/2$ - we know the integral converges, since $f\in C$. Thus $\forall s>0$, we can choose an $x(s)$ close to 1, s.t. that lhs is smaller than $s$. This follows directly from the convergence i.e. the continuity of $F$. – Diger Mar 25 '23 at 19:45