2

Have difficulty in understanding the proof below, in the book, by Traynor, titled: Introduction to analysis, at page #27; with title: The existence of roots — a consequence of completeness.

The proof is stated to show that irrational roots of $\sqrt{2}$ exist.

Theorem. Let $y$ be a positive real number. Then, for every $n\in \mathbb{N}$, there exists a unique positive real number such that $x^n = y$.

Proof.

(1) First, note that for positive numbers $a, b,\,\, a\lt b \implies a^n < b^n$.
This is proved by induction on $n$. (Exercise).

(2) This implies uniqueness: $\,\,$Suppose $x_1^n = y\,\,$ and $\,\,x_2^n = y\,\,$, with $\,\,x_1, x_2\,\,$ positive but not equal. Then one must be smaller, by trichotomy. Say $\,\,x_1\lt x_2$.
Then $\,\,x_1^n \lt x_2^n$, so we can't have $\,\,x_1^n = x_2^n$. The contradiction shows $\,\,x_1 = x_2$.

(3) Let $A = \{a \gt 0 : a^n \le y\}, B = \{b \gt 0 : b^n \ge y\}$. I claim that $A$ and $B$ are not empty and every element of $A$ is $\le$ every element of $B$.

Indeed, since $y \gt 0, \,\, 0 \lt \frac y{y+1}\lt 1$, so we have $$0 \lt (\frac {y}{y+1})^n \le \frac y{y+1}\lt y$$.

Thus, $\frac y{y+1}\in A$. In the same way, $y + 1 \gt 1$, so $(y + 1)^n \ge y + 1 \gt y$ and hence $y + 1 \in B$.

Now, if $a \in A$ and $b \in B$ we have $a^n \le y \le b^n $, so $a^n \le b^n$, and therefore $a \le b$.
This comes from step 1, because if $a\gt b$ we would have $a^n \gt b^n$.

(4) Step 3 sets us up for our completeness axiom.

There must exist an $x$ with $a\le x \le b$, for all $a \in A$ and all $b\in B$. We will now show that for this $x, \,\, x^n = y$.

Let $\, 0 \lt a\lt x$. Then, $\, a \in A$.
For suppose not; then, $\, a^n \gt y$, which makes $\, a \in B$ and $\, x \le$ each element of $B$, a contradiction.
Similarly, if we let $\,b\gt x \,$, then $\, b \in B$.

Thus, $$a^n \lt x^n \lt b^n, \,\,\,\,\,\,a^n \le y \le b^n.$$

If we multiply the second string of inequalities by $\,−1\,$, they turn around, so $$a^n \lt x^n \lt b^n\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,$$ $$\,\,\,\,−b^n \le −y \le −a^n\,\,\,\, [Prob. 1$$ and adding gives $$\,a^n − b^n \lt x^n − y\lt \,b^n − a^n.\,\,\,[Prob. 2$$ In other words, $$|x^n − y| \le b^n − a^n \le (b − a)nb^{n−1}.\,\,\,[Prob. 3$$ Here we used the fact that $$b^n − a^n = (b − a)\sum_{i=1}^{n}b^{n-i} a^{i-1}\le (b − a)nb^{n−1}$$ Now take any $\epsilon$ such that $\,0 <\epsilon \lt x, \,a = x − \epsilon, \, b = x + \epsilon$.
Then $$\,b − a < 2\epsilon\,,\,b < 2x \,\,\,\, [Prob. 4 $$ and so $$|x^n − y| \le 2\,\epsilon \, n(2x)^{n−1}.$$ Hence, $$0 ≤ \frac{|x^n − y|}{2n(2x)^{n−1}} \le \epsilon.$$ Now $\,\epsilon \,\,$ here was arbitrary satisfying $\,0 \lt \epsilon\lt x$.
So, $$\frac{|x^n − y|}{2n(2x)^{n−1}} = 0.$$ Hence $|x^n −y| = 0$. But then $x^n −y = 0$, so $x^n = y$, which completes the proof.

enter image description here

The details of the problems problem numbers are stated below.

Prob. 1: Unable to show how multiplying $a^n \lt x^n \lt b^n$ by $-1$ leads to $−b^n \le −y \le −a^n$.

Prob. 2: How the equality sign gets removed upon adding the two inequalities?

Prob. 3: How the equality sign is re-introduced when modulus is taken of $|x^n − y|$?

Prob. 4: It seems like a typographical error is there in stating $\,\,b − a\lt 2\epsilon$.

$a = x − \epsilon, \, b = x + \epsilon$. Then $\,\,b − a = 2\epsilon\,\,$ & not $\,\lt 2\epsilon$.
Also, the correct ordering (equality, instead of less than) still brings the correct relation : $b < 2x$, as $\,\,b − a = 2\epsilon\implies b = x+\epsilon\implies b \lt x+x$, as $x\gt \epsilon$.

Also, have a feeling that the proof is not rigorous enough, as relies on so many assumptions and so request another method which is more rigorous for the same.

jiten
  • 4,524

2 Answers2

2

Problem 1: This is a basic theorem of ordered fields: If $a \le b$, then $-a \ge -b$. Proof: \begin{align*} a \le b &\implies -a + a \le -a + b \\ &\implies 0 \le -a + b \\ &\implies 0 + (-b) \le -a + b + (-b) \\ &\implies 0 + (-b) \le -a + 0 \\ &\implies -b \le -a. \end{align*} The given version is the double inequality version of this. Saying $a^n \le y \le b^n$ means $a^n \le y$ and $y \le b^n$. The above result implies $-y \le -a^n$ and $-b^n \le -y$, i.e. $$-b^n \le -y \le a^n.$$

Problem 2: Another basic theorem: if $a \le b$ and $c \le d$, then $a + c \le b + d$. Proof: \begin{align*} (a \le b) \land (c \le d) &\implies (a + c \le b + c) \land (b + c \le b + d) \\ &\implies a + c \le b + d. \end{align*} Now, let us suppose that $a + c = b + d$. This would mean (by trichotomy) that $$a + c \le b + c \le b + d = a + c \implies b + c = a + c = b + d.$$ Using cancellation law, this implies $a = b$ and $c = d$. So, if we have $a < b$ (i.e. $a \neq b$ as an additional assumption) and $c \le d$, then we naturally have $a + c < b + d$. This is why the equal sign has disappeared.

Problem 3: The equality sign shows only the possibility of equality. We can always add it back in, as it only weakens the statement. That is, we always have $a < b \implies a \le b$ (and not always the other way around).

As to why they've chosen to add the possibility of equality back in, I'm not sure. It doesn't add (nor does it detract) from the proof.

Problem 4: Agreed; it should be $b - a = 2\varepsilon$. However, this appears to be the only mistake in the proof. Aside from this, and Lord Shark's accurate comment that the Intermediate Value Theorem would make quicker, more elegant work of this, the proof seems fine.

Theo Bendit
  • 50,900
  • 4
    In the context of the book the IVT has not been shown at all. The proof wants to only use algebra and order /completeness, no more. It's not about giving the slickest proof possible. And the IVT also follows from connectedness which is a consequence of completeness (and order denseness) so is not really a simpler proof. – Henno Brandsma Jun 17 '19 at 05:57
  • 2
    @HennoBrandsma I would say that it is a simpler way of going about proving this. Essentially, the proof as written is really mashing two unrelated steps in a single proof: it's proving the continuity of the $a \mapsto a^n$ map, and proving a special case of the IVT. I think the more elegant (and sensible) approach is to wait until the IVT is proven, prove that $a \mapsto a^n$ is continuous, and have a two line corollary saying that $n$th roots exist. The proof is long-winded enough that the asker seemed dubious enough about its validity. Structuring the theory better would rectify this. – Theo Bendit Jun 17 '19 at 06:03
  • In the post, it is given that $a^n \lt x^n \lt b^n$, $a^n \le y \le b^n$. Why equality is given for $y$ and not for $x^n$. The whole confusion stems from this alone. – jiten Jun 17 '19 at 06:09
  • @HennoBrandsma Please elaborate your comment so that can understand better the intent. Particularly, the last line, and what is being referred to as 'not really a simpler proof'. – jiten Jun 17 '19 at 06:13
  • 1
    @jiten Remember that $a$ and $b$ were initially chosen to satisfy $0 < a < x$ and $x < b$, and so $a^n < x^n < b^n$. The proof then concludes $a \in A$ and $b \in B$ (check the proof to make sure you're ok with these arguments). The definition of $a \in A$ means $a^n \le y$ and $b \in B$ means $y \le b^n$; check back where $A$ and $B$ are defined! That said, most of the strict inequalities here could be made non-strict, and the proof would still be valid. – Theo Bendit Jun 17 '19 at 06:15
  • Please explain in detail your line in response to comment by @HennoBrandsma: 'Essentially, the proof as written is really mashing two unrelated steps in a single proof: it's proving the continuity of the $a\rightarrow a^n$ map, and proving a special case of the IVT.' – jiten Jun 17 '19 at 07:00
  • 1
    @jiten Well, the proof of the Intermediate Value Theorem (the one I'm familiar with, at least) involves a set very much like $A$, or a set very much like $B$. If you have $a < b$ and $f(a) < c < f(b)$, where $f : [a, b] \to \Bbb{R}$ is continuous, you can define $A = {x \in [a, b] : f(x) \le c}$. The supremum of this set (guaranteed by completeness) will have to satisfy $f(x) = c$ by continuity. This is essentially the IVT in a nutshell. This is more or less what is happening in the above proof. – Theo Bendit Jun 17 '19 at 07:05
  • 1
    @jiten The continuity of $a \mapsto a^n$ is, of course, a vital ingredient, and the given proof includes this too, in the form of the second half of the fourth step. – Theo Bendit Jun 17 '19 at 07:06
  • Thanks a lot. But, am still unable to find the proof for $a\rightarrow a^n$ in the second part of the fourth step. Sorry, if naive. – jiten Jun 17 '19 at 07:20
  • 1
    @jiten The part I'm referring to is the part that begins "Now take any $\epsilon$ such that...". It's not continuity in the usual sense, but it's the analysis argument that steps in for continuity. Proving continuity of $a \mapsto a^n$ follows a very similar trajectory, involving similar inequalities. In particular, the inequalities provided imply that, given $a, b \in [0, M]$, we have that $|b^n - a^n| \le n M^{n - 1}|b - a|$, making $a \mapsto a^n$ locally Lipschitz, a stronger form of continuity. – Theo Bendit Jun 17 '19 at 07:43
  • @jiten I wouldn't worry if you're not following this stuff in the comments though. You should probably wait until at least you've covered the IVT before you try to understand this stuff properly. – Theo Bendit Jun 17 '19 at 07:45
1

For problem 1, the multiplication is not applied to $a^n<x^n<b^b$ but rather $a^n < y < b^n$.

For problem 2, if we have $a<b$ and $c \le d$, consider two cases.

If $c=d$, then we have $a+c< b+c$ that is $a+c < b+d$.

If $c<d$, then we have $a+c< b +d$.

For problem $3$, you don't have to include the equality but there is nothing wrong to include it.

Yes, there is a typo. The quickest fix is just let $b-a=2\epsilon$. Alternatively, you can also choose $x-\epsilon \le a<x$ and $x < b \le x+\epsilon$.

You are building a structure from the axioms. Here you have very little results to use. You have not built intermediate value theorem or even continuity yet. You are supposed to only use results that have been stated prior to proving this. Mainly axioms for the real number as stated in the book (field axioms, order axioms, and completeness axiom). Hence the construction of $A$ and $B$ in order to use the completeness axiom.

Siong Thye Goh
  • 149,520
  • 20
  • 88
  • 149
  • Thanks, but am unable to understand why the continuity of $a\rightarrow a^n$ is shown in the proof. If you need the reference of the book, it is not available anywhere online or for offline purchase, not even on author's site, but at: https://b-ok.org/book/465886/8c35b2. Do you feel the author hints about this in any detail at the book's page #27, or later or earlier. If free, then please discuss it at the older chat room at: https://chat.stackexchange.com/rooms/94404/discussion-between-siong-thye-goh-and-jiten. – jiten Jun 18 '19 at 07:17
  • I am very confused over my inability to solve the issues sated in my last comment. Please help. – jiten Jun 18 '19 at 08:42
  • 1
    $b^n − a^n = (b − a)\sum_{i=1}^{n}b^{n-i} a^{i-1}\le (b − a)nb^{n−1}$ implies continuity. – Siong Thye Goh Jun 18 '19 at 08:45
  • 1
    I am not clear how $b^n − a^n = (b − a)\sum_{i=1}^{n}b^{n-i} a^{i-1}\le (b − a)nb^{n−1}$ implies continuity. Request more inputs. – jiten Jun 18 '19 at 08:54
  • hmmm... do u know what is continuity in the first place? – Siong Thye Goh Jun 18 '19 at 09:12
  • 1
    sorry, was not there. continuity means that there exists a value of $f(x),,, \forall x$ in a given interval. It is different from being differentiable, which is a stronger criteria. – jiten Jun 18 '19 at 09:42
  • Nope, your definition of continuity is not correct. – Siong Thye Goh Jun 18 '19 at 09:57
  • Am unable to get correct defn. except in 2 places: 1) book on analysis by Denlinger, 2) site at : http://oregonstate.edu/instruct/mth251/Jwlee/cq/Stage4/Lesson/continuity.html;
    states> We are looking for a mathl. defn. which captures two ideas.
    • The values of fn. $f(x)$ at points near $a$ are good predictors of the value of $f$ at $a$.
    • The graph of $f$ is a connected curve with no jumps, gaps, or holes. Also, does it mean: left hand limit = right hand limit, apart from limit existing at all points in the interval.
    – jiten Jun 19 '19 at 18:18
  • Please be in chat for explaining your comment ($b^n − a^n = (b − a)\sum_{i=1}^{n}b^{n-i} a^{i-1}\le (b − a)nb^{n−1}$ ) as it is taking too much time for me. I also need why you stated that as implying continuity, when I asked about why continuity of $a\rightarrow a^n$ is shown. The older chatroom is at:https://chat.stackexchange.com/rooms/94404/discussion-between-siong-thye-goh-and-jiten. – jiten Jun 21 '19 at 13:07
  • Please help with my function transformation post at : https://math.stackexchange.com/q/3270550/424260. – jiten Jun 22 '19 at 10:54
  • I hope your help will be forthcoming. – jiten Jun 22 '19 at 11:35