Some Properties of the Rationals .... Bloch Ex. 1.5.9 (3)

[Math Amateur]

Homework Statement

I am reading Ethan D. Bloch's book: The Real Numbers and Real Analysis ...

I am currently focused on Section 1.5: Constructing the Rational Numbers ...

I need help with Exercise 1.5.9 (3) ...Exercise 1.5.9 reads as follows:

Now ... we wish to prove that for $r, s \in \mathbb{Q}$ where $r \gt 0$ and $s \gt 0$ that:

If $r^2 \lt s$ then there is some $k \in \mathbb{N}$ such that $( r + \frac{1}{k} )^2 \lt s$ ... ...

Homework Equations

... and relevant information ...[/B]

We are at the point in Bloch's book where he has just defined/constructed the rational numbers, having previously defined/constructed the natural numbers and the integers ... so (I imagine) at this point we cannot assume the existence of the real numbers.

Basically Bloch has defined/constructed the rational numbers as a set of equivalence classes on $\mathbb{Z} \times \mathbb{Z}^*$ and then has proved the usual fundamental algebraic properties of the rationals ...

The Attempt at a Solution

Solution Strategy

Prove that there exists a $k \in \mathbb{N}$ such that $( r + \frac{1}{k} )^2 \lt s$ ... BUT ... without in the proof involving real numbers like $\sqrt{2}$ because we have only defined/constructed $\mathbb{N}, \mathbb{Z}$, and $\mathbb{Q}$ ... so I am assuming that we cannot take the square root of the relation $( r + \frac{1}{k} )^2 \lt s$ and start dealing with a quantity like $\sqrt{s}$ ... is this a sensible assumption ...?So ... assume $( r + \frac{1}{k} )^2 \lt s$ ..

then

$( r + \frac{1}{k} )^2 \lt s$

$\Longrightarrow r^2 + \frac{2r}{k} + \frac{1}{k^2} \lt s$

$\Longrightarrow r^2 + \frac{1}{k^2} \lt s$ ... ... since $\frac{2r}{k} \gt 0$ ... (but ... how do I justify this step?)

$\Longrightarrow k^2 \gt \frac{1}{ s - r^2 }$

But where do we go from here ... seems intuitively that such a $k \in \mathbb{N}$ exists ... but how do we prove it ...

(Note that I am assuming that for $k \in \mathbb{N}$ that if we show that $k^2$ exists, then we know that $k$ exists ... is that correct?Hope that someone can clarify the above ...

Help will be much appreciated ...

Peter

===========================================================================================

***NOTE***

In Exercises 1.5.6 to 1.5.8 Bloch gives a series of relations/formulas that may be useful in proving Exercise 1.5.9 (indeed, 1.5.9 (1) and (2) may be useful as well) ... so I am providing Exercises 1.5.6 to 1.5.8 as follows: (for 1.5.9 (1) and (2) please see above)

 

[andrewkirk]

As is often the case in problems like this, one uses earlier parts to prove later parts. We can do that if we can turn the problem into one of searching for a $n$ that solves a problem of type (1) and a $m$ that solves a problem of type (2).

Very broadly speaking, we want to convert it into a problem of showing that certain rational numbers are smaller than some other rational number.

We have been asked to show there exists $k$ such that
$$(r+\frac1k)^2<s$$
One of the first things to try in any such situation is to expand a product. Doing that on the left allows us to restate the target inequality as
$$r^2 +\frac {2r}k +\frac1{k^2}<s$$
But we also need to use what we've been given, which is that $r^2<s$. In other words $h=s-r^2$ is positive, and rational.

We can use that to rewrite what we need to prove as (there exists $k\in\mathbb N$ such that)
$$\frac {2r}k +\frac1{k^2}<h\quad\quad\quad\quad\quad\quad(\textrm{i})$$

Rather than continue, and spoil your fun, can you think of a way to convert that into a problem of type (1) and another of type (2)? A minor obstacle is that we have only one inequality, and we need two, to get two problems. But we know in our guts that the inequality must be true and that we have lots of room to move, so we can adopt two inequalities that are sufficient but not necessary conditions for the above inequality to hold, and then use those two inequalities in parts (1) and (2).

Can you think of two inequalities that, if both true, guarantee that (i) is true?

EDIT: I just realized that, if you can prove that $k\in\mathbb N\to \frac1k \leq1$ then it can be solved in one part rather than two, using only (1).

 

[haruspex]

If you are trying to prove something that depends on the variables being rational numbers then it is a good idea to invoke the definition of a rational number at some point. If r is a rational then there exist ...

 

[Math Amateur]

As is often the case in problems like this, one uses earlier parts to prove later parts. We can do that if we can turn the problem into one of searching for a $n$ that solves a problem of type (1) and a $m$ that solves a problem of type (2).

Very broadly speaking, we want to convert it into a problem of showing that certain rational numbers are smaller than some other rational number.

We have been asked to show there exists $k$ such that
$$(r+\frac1k)^2<s$$
One of the first things to try in any such situation is to expand a product. Doing that on the left allows us to restate the target inequality as
$$r^2 +\frac {2r}k +\frac1{k^2}<s$$
But we also need to use what we've been given, which is that $r^2<s$. In other words $h=s-r^2$ is positive, and rational.

We can use that to rewrite what we need to prove as (there exists $k\in\mathbb N$ such that)
$$\frac {2r}k +\frac1{k^2}<h\quad\quad\quad\quad\quad\quad(\textrm{i})$$

Rather than continue, and spoil your fun, can you think of a way to convert that into a problem of type (1) and another of type (2)? A minor obstacle is that we have only one inequality, and we need two, to get two problems. But we know in our guts that the inequality must be true and that we have lots of room to move, so we can adopt two inequalities that are sufficient but not necessary conditions for the above inequality to hold, and then use those two inequalities in parts (1) and (2).

Can you think of two inequalities that, if both true, guarantee that (i) is true?

EDIT: I just realized that, if you can prove that $k\in\mathbb N\to \frac1k \leq1$ then it can be solved in one part rather than two, using only (1).

Hi Andrew ... thanks for the help!

I take it that by problem of type (1) and type (2) you mean Exercise 1.5.9 (1) and Exercise 1.5.9 (2) ... ... I think you do mean this ... but just checking ...

Peter

 

[Math Amateur]

If you are trying to prove something that depends on the variables being rational numbers then it is a good idea to invoke the definition of a rational number at some point. If r is a rational then there exist ...

Hi haruspex ... yes, indeed you are right ...

Bloch, through a series of definitions and theorems shows us that we can legitimately think of rational numbers as expressions of the form ( "fractions" ) $\frac{a}{b}$ where $a, b \in \mathbb{Z}$ and where $b \ne 0$ ... ... and also proves that we may manipulate these entities as we did in high school ... I was taking this idea of "fractions" as my working definition of rationals ... but should have stated it ... mind you Bloch takes 5 pages of his text to show that this is legitimate ...

Peter

 

[haruspex]

Hi haruspex ... yes, indeed you are right ...

Bloch, through a series of definitions and theorems shows us that we can legitimately think of rational numbers as expressions of the form ( "fractions" ) $\frac{a}{b}$ where $a, b \in \mathbb{Z}$ and where $b \ne 0$ ... ... and also proves that we may manipulate these entities as we did in high school ... I was taking this idea of "fractions" as my working definition of rationals ... but should have stated it ... mind you Bloch takes 5 pages of his text to show that this is legitimate ...

Peter

Ok, so complete the sentence

If r is a rational then there exist ...

 

[Math Amateur]

As is often the case in problems like this, one uses earlier parts to prove later parts. We can do that if we can turn the problem into one of searching for a $n$ that solves a problem of type (1) and a $m$ that solves a problem of type (2).

Very broadly speaking, we want to convert it into a problem of showing that certain rational numbers are smaller than some other rational number.

We have been asked to show there exists $k$ such that
$$(r+\frac1k)^2<s$$
One of the first things to try in any such situation is to expand a product. Doing that on the left allows us to restate the target inequality as
$$r^2 +\frac {2r}k +\frac1{k^2}<s$$
But we also need to use what we've been given, which is that $r^2<s$. In other words $h=s-r^2$ is positive, and rational.

We can use that to rewrite what we need to prove as (there exists $k\in\mathbb N$ such that)
$$\frac {2r}k +\frac1{k^2}<h\quad\quad\quad\quad\quad\quad(\textrm{i})$$

Rather than continue, and spoil your fun, can you think of a way to convert that into a problem of type (1) and another of type (2)? A minor obstacle is that we have only one inequality, and we need two, to get two problems. But we know in our guts that the inequality must be true and that we have lots of room to move, so we can adopt two inequalities that are sufficient but not necessary conditions for the above inequality to hold, and then use those two inequalities in parts (1) and (2).

Can you think of two inequalities that, if both true, guarantee that (i) is true?

EDIT: I just realized that, if you can prove that $k\in\mathbb N\to \frac1k \leq1$ then it can be solved in one part rather than two, using only (1).

Thanks again, Andrew ...

Think that this may be one way to go ...

We have $\frac{2r}{k} + \frac{1}{k^2} \lt h$ where $r, h \in \mathbb{Q}, \ k \in \mathbb{N}, \ r, h \gt 0$

Now ...

$\frac{2r}{k} + \frac{1}{k^2} \lt h$

$\Longrightarrow 2rk + 1 \lt h k^2$ ... (Multiply through by k^2 )

$\Longrightarrow 2rk \lt h k^2$

$\Longrightarrow 2 r \lt h k$ ... ... (Divide through by k)

Since $2r \gt 0$ and $h \gt 0$ we can use result in 1.5.9 (1) with $2r = s$ and $k = n$ and $h = r$ ...

... ... then we have that there exists a $k$ such that $2r \lt k h$ ...Is that correct?

PeterNote: Andrew ... interested to know how you used both (1) and (2) ... I could not see how to do that ..

 

[Math Amateur]

Ok, so complete the sentence

To complete the sentence ...

If $r$ is a rational then there exist $a, b \in \mathbb{Z}$ and $\mathbb{Z}^*$ respectively such that $r = \frac{a}{b}$ ...

Hope that is correct ...

Peter

 

[haruspex]

To complete the sentence ...

If $r$ is a rational then there exist $a, b \in \mathbb{Z}$ and $\mathbb{Z}^*$ respectively such that $r = \frac{a}{b}$ ...

Hope that is correct ...

Peter

Ok. So do the same with s and use these to replace r and s in the given relationship (r²<s).

 

[andrewkirk]

We have $\frac{2r}{k} + \frac{1}{k^2} \lt h$ where $r, h \in \mathbb{Q}, \ k \in \mathbb{N}, \ r, h \gt 0$

Now ...

$\frac{2r}{k} + \frac{1}{k^2} \lt h$

$\Longrightarrow 2rk + 1 \lt h k^2$ ... (Multiply through by k^2 )

$\Longrightarrow 2rk \lt h k^2$

$\Longrightarrow 2 r \lt h k$ ... ... (Divide through by k)

Since $2r \gt 0$ and $h \gt 0$ we can use result in 1.5.9 (1) with $2r = s$ and $k = n$ and $h = r$ ...

... ... then we have that there exists a $k$ such that $2r \lt k h$ ...Is that correct?

Almost, but not quite. The arrows need to go the other way, because it is at the last bit that you use (1) to show that a $k$ exists that has certain properties, then, working backwards through the above (with reversed arrows) we prove that it also has the property of satisfying the original inequality.

I usually do these problems the way you have, working forwards, but then checking that I can reverse it at the end, which relies on the arrows all being double-ended ($\Leftrightarrow$).

Or we can do it in one pass, by requiring all the arrows to point in the other direction.

There's probably often a better way, but I usually don't find it.

Working from the bottom up, we start with
$$\exists k\in \mathbb N:\ 2r<kh$$
Given that $k>0$, that is logically equivalent to ($\Leftrightarrow$)
$$\exists k\in \mathbb N:\ 2rk<hk^2$$
But then we strike a snag. Because we cannot reverse the next arrow. That is
$$(\exists k\in \mathbb N:\ 2rk<hk^2)
\not\Rightarrow
(\exists k\in \mathbb N:\ 2rk+1<hk^2)$$

Let's try to fix that:

Taking $\mathbb N$ to exclude 0 so that $k\in\mathbb N\Rightarrow k>0$ we have
\begin{align*}
\exists k\in\mathbb N:\ k\in\mathbb N:\ \left(r+\frac1k\right)^2<s &\Leftrightarrow
\exists k\in\mathbb N:\ r^2 + \frac{2r}k +\frac1{k^2}<s \\
&\Leftrightarrow\ \
\exists k\in\mathbb N:\ \frac{2r}{k} + \frac{1}{k^2} < h\\
&\Leftrightarrow\ \
\exists k\in\mathbb N:\ 2rk + 1 < hk^2
\end{align*}

That's where we previously hit the snag in reversing. We need to connect to a new inequality via a left-arrow $\Leftarrow$, where the new inequality is something we can still achieve. Removing the $1$ won't do it. We need to make the LHS bigger. So instead let's replace $1$ by $k$. Then our next step is
$$\Leftarrow
\exists k\in\mathbb N:\ 2rk + k < hk^2$$
provided we have $k\in\mathbb N\Rightarrow k\geq 1$. Do you have that given? If not we'll need to use a different trick.
Assuming we have that, next we can divide by $k$ to get
$$\ \ \Leftrightarrow\ \
\exists k\in\mathbb N:\ 2r+1<kh$$
and we can use (1) to show that last one is true.

Then we can start with the last inequality, known to be true, and follow the left-pointing arrows to get to the original inequality that we wanted to prove.

BTW, my approach using (1) and (2) was to go from
\begin{align*}
\exists k\in\mathbb N:\ k\in\mathbb N:\ \left(r+\frac1k\right)^2<s &\Leftrightarrow
\exists k\in\mathbb N:\ r^2 + \frac{2r}k +\frac1{k^2}<s \\
&\Leftrightarrow\ \
\exists k\in\mathbb N:\ \frac{2r}{k} + \frac{1}{k^2} < h\\
&\Leftarrow\ \
\exists k\in\mathbb N:\bigg(\left(\frac{2r}k<\frac h2\right)\wedge
\left(\exists k\in\mathbb N:\ \frac1{k^2}<\frac h2\right)\bigg)
\end{align*}
Then we use (1) to prove the first conjunct and (2) to prove the second, then take the greater of the two $k$s from the two conjuncts.

As the edit in my post shows, that is not necessary here, but it's a trick that is often useful in other analytical proofs - splitting a tolerance margin ($h$ in this case) into sub-margins, proving a separate inequality for each one, then showing we can choose a number ($k$ in this case) that satisfies them all simultaneously.

 

[Math Amateur]

Ok. So do the same with s and use these to replace r and s in the given relationship (r²<s).

To formulate the question as you wish we would proceed as follows:

Let $r = \frac{a}{b}$ and let $s = \frac{c}{d}$ ...

Then we have $\frac{a}{b} \gt 0$ and $\frac{c}{d} \gt 0$ where $a,b,c,d \in \mathbb{Z}$ ...

and we can assume that $a, b, c, d \gt 0$ ... ( ... if $r = \frac{a}{b}$ where $a, b$ both negative then multiply $a, b$ by $-1$ and redefine $a, b$ ... )

We assume $r^2 \lt s$ ... that is we assume $\frac{a^2}{b^2} \lt \frac{c}{d}$ ... ...

We have to show that given the above there exists a $k \in \mathbb{N}$ such that:

$( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ ...

If we follow the same strategy as Andrew followed then we would investigate the conditions for there to exist a $k$ such that $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ hoping to find a condition ... being careful to establish a chain of 'double' or two-way implications ... and then use the reverse implication chain to establish a proof ...

Peter

 

[haruspex]

that is we assume $\frac{a^2}{b^2} \lt \frac{c}{d}$

Ok. Multiply that out. Does the resulting inequality suggest anything?

 

[Math Amateur]

Ok. Multiply that out. Does the resulting inequality suggest anything?

$\frac{a^2}{b^2} \lt \frac{c}{c} \ \Longleftrightarrow a^2 d \lt b^2 c$ ... ... Now ... obviously from what you have said ... that should suggest something to me ... but ... no ideas ...

Sorry ... ... can you help further ...

Peter

 

[haruspex]

$\frac{a^2}{b^2} \lt \frac{c}{c} \ \Longleftrightarrow a^2 d \lt b^2 c$ ... ...

What friendly sort of beast do you have on each side of $a^2 d \lt b^2 c$? What species of mathematical animal were a, b, c and d defined to be?

 

[Math Amateur]

What friendly sort of beast do you have on each side of $a^2 d \lt b^2 c$? What species of mathematical animal were a, b, c and d defined to be?

... well ... both $a^2 d$ and $b^2 c$ are positive integers or natural numbers ... as are $a, b, c$ and $d$ ... ...

Peter

 

[haruspex]

... well ... both $a^2 d$ and $b^2 c$ are positive integers or natural numbers ...

Peter

Right. And if integer m < integer n, can you say something just a bit stronger about their relationship?

 

[Math Amateur]

Right. And if integer m < integer n, can you say something just a bit stronger about their relationship?

... For $m, n \in \mathbb{Z}$ then ... ... If $m \lt n$ then there exists $p \in \mathbb{Z}$ such that $m + p = n$ ... ...

Peter

 

[haruspex]

... For $m, n \in \mathbb{Z}$ then ... ... If $m \lt n$ then there exists $p \in \mathbb{Z}$ such that $m + p = n$ ... ...

Peter

How near to being equal, while not being equal, can two integers be?

 

[Math Amateur]

How near to being equal, while not being equal, can two integers be?

If $m,n$ are as near equal as two integers can get then $m + 1 = n$ ... ...

Peter

 

[haruspex]

If $m,n$ are as near equal as two integers can get then $m + 1 = n$ ... ...

Peter

(Or m = n+1)
Right, so what inequality can you write that is a bit stronger than just m < n?

 

[Math Amateur]

(Or m = n+1)
Right, so what inequality can you write that is a bit stronger than just m < n?

Well ... you could write $m + 1 \le n$ ... is that what you are getting at?

Peter

 

[haruspex]

Well ... you could write $m + 1 \le n$ ... is that what you are getting at?

Peter

Yes.
The point is that there is no limit to how close two rationals can be, but there is this limit on how close two integers can be. And it is this window that gives you a way to squeeze something in between in a predictable way.
Can you now find a value of k that achieves what you need?

 

[Math Amateur]

Yes.
The point is that there is no limit to how close two rationals can be, but there is this limit on how close two integers can be. And it is this window that gives you a way to squeeze something in between in a predictable way.
Can you now find a value of k that achieves what you need?

Would a sensible start to deriving $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ for some $k \in \mathbb{N}$ be as follows:

$\frac{a^2}{b^2} \lt \frac{c}{d}$

$\Longrightarrow a^2 d \lt b^2 c$

$\Longrightarrow a^2 d + 1 \le b^2 c$

$\Longrightarrow a^2 d + \frac{1}{k^2} \lt b^2 c$

$\Longrightarrow \frac{a^2 d}{b^2} + \frac{1}{k^2 b^2} \lt c$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{k^2 b^2 d} \lt \frac{c}{d}$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{k^2 b^2 d^2} \lt \frac{c}{d}$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{m^2} \lt \frac{c}{d}$

where $m \in \mathbb{N}$
But how do I amend and complete this proof ... it misses the mark a bit ...

... the middle term of the square $( \frac{a}{b} + \frac{1}{m} )^2$ namely $\frac{2ab}{m }$ needs to be added to LHS above ... but ... problem ...

Peter

 

[haruspex]

Would a sensible start to deriving $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ for some $k \in \mathbb{N}$ be as follows:

$\frac{a^2}{b^2} \lt \frac{c}{d}$

$\Longrightarrow a^2 d \lt b^2 c$

$\Longrightarrow a^2 d + 1 \le b^2 c$

$\Longrightarrow a^2 d + \frac{1}{k^2} \lt b^2 c$

$\Longrightarrow \frac{a^2 d}{b^2} + \frac{1}{k^2 b^2} \lt c$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{k^2 b^2 d} \lt \frac{c}{d}$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{k^2 b^2 d^2} \lt \frac{c}{d}$

$\Longrightarrow \frac{a^2 }{b^2} + \frac{1}{m^2} \lt \frac{c}{d}$

where $m \in \mathbb{N}$
But how do I amend and complete this proof ... it misses the mark a bit ...

Peter

Expand the (...)² term and multiply out the same way as before to get rid of the fractions.

 

[Math Amateur]

Expand the (...)² term and multiply out the same way as before to get rid of the fractions.

Do you mean start from the inequality $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ and expand its squared term ... and work from there using the approach that Andrew took ... a sort of indirect approach ...

Peter

 

[haruspex]

Do you mean start from the inequality $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ and expand its squared term ...

Yes.
Sorry, have to offline for a few hours.

 

[Math Amateur]

Yes.
Sorry, have to offline for a few hours.

OK ... thanks so much for the help ...

Will watch for your return

Thanks again,

Peter

 

[Math Amateur]

Yes.
Sorry, have to offline for a few hours.

So if we start from $( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ ... ...

... ... we have ...

$( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$

$\Longleftrightarrow ( \frac{a^2}{b^2} + \frac{2a}{bk} + \frac{1}{k^2} ) \lt \frac{c}{d}$

$\Longleftrightarrow a^2 k^2 d + 2a b d k + b^2 d \lt c b^2 k^2$

... ... but ... ... how do we proceed from here ...

... presumably we use the given inequality, namely $a^2 d \lt c b^2$ ... ... but how, exactly ... Can you help me to proceed from here ...?

Peter

 

[haruspex]

presumably we use the given inequality, namely a²d<cb²

No, we use the tighter form, a²d+1≤cb². You need to find a value of k such that it follows from that that

a²k²d+2abdk+b²d<cb²k²

 

[Math Amateur]

No, we use the tighter form, a²d+1≤cb². You need to find a value of k such that it follows from that that

So ... we need to simplify

$a^2k^2d + 2abdk + b^2d \lt cb^2k^2$ ... ... ... ... (1)

so that we show that $k \in \mathbb{N}$ exists, then use the reverse implications to show that $k$ satisfies the following inequality:

$( \frac{a}{b} + \frac{1}{k} )^2 \lt \frac{c}{d}$ ... ... ... ... (2)

Now we are given that $\frac{a^2}{b^2} \lt \frac{c}{d}$ which gives us the following inequality:

$a^2d + 1 \le cb^2$ ... ... ... ... (3)

And if we multiply through (3) by $k^2$ we get the following:

$a^2dk^2 + k^2 \le cb^2k^2$ ... ... ... ... (4)

Now I note that the terms $a^2dk^2$ and $cb^2k^2$ from (4) also appear in (1) ... but I cannot see how to use this ... or anything else to proceed forward ...

... can you help further ...

Peter

 

[haruspex]

I note that the terms $a^2dk^2$ and $cb^2k^2$ from (4) also appear in (1)

Yes, this is key.
It might help clarify matters if you put those terms on the left and all else on the right in each inequality.
Be careful to keep track of which inequality is known to be true and which you are trying to deduce.

 

[Math Amateur]

Yes, this is key.
It might help clarify matters if you put those terms on the left and all else on the right in each inequality.
Be careful to keep track of which inequality is known to be true and which you are trying to deduce.

Following your advice ... we have ...

$(1) \Longleftrightarrow a^2k^2d - cb^2k^2 \lt -2abdk - b^2d$ ... ... ... ... (5)

and

$(4) \Longleftrightarrow a^2k^2d - cb^2k^2 \le -k^2$ ... ... ... ... (6)(6) is true by assumption and we are trying to find a condition on $k \in \mathbb(N)$ from (5)BUT ... what step do I take now ... still perplexed as to how to proceed ...

Can you help ... ...?

Peter

 

[haruspex]

we are trying to find a condition on $k \in \mathbb(N)$ from (5)

You need to come up with a choice for k such that if (6) is true then (5) must be true.
Do you think that will be a large value of k or a small one?

 

[Math Amateur]

You need to come up with a choice for k such that if (6) is true then (5) must be true.
Do you think that will be a large value of k or a small one?

Well ... the size of k doesn't matter in (6) ... .. but in (5) it would be more likely true if k were large ,,, is that right?

Peter

 

[haruspex]

the size of k doesn't matter in (6)

It does. It might help to focus by calling the terms on the left of the inequalities just "LHS", as an opaque package.
As k increases, what does (6) tell you is happening to the LHS? Compare that with what happens as k increases in (5).