5

In Spivak's calculus book, I cannot understand the solution proposed for question (c) of problem 18 in chapter 2:

Prove that $\sqrt{2}+\sqrt[3]{2}$ is irrational. Hint: Start by working out the first 6 powers of this number.

Working out the powers is quite easy:

  • $(2^\frac{3}{6}+2^\frac{2}{6})^0 = 1$
  • $(2^\frac{3}{6}+2^\frac{2}{6})^1 = 2^\frac{3}{6} + 2^\frac{2}{6}$
  • $(2^\frac{3}{6}+2^\frac{2}{6})^2 = 2^\frac{6}{6} + 2 \cdot 2^\frac{3}{6} \cdot 2^\frac{2}{6} + 2^\frac{4}{6} = \\ 2 \cdot 2^\frac{0}{6} + 2^\frac{4}{6} + 2 \cdot 2^\frac{5}{6}$
  • $(2^\frac{3}{6}+2^\frac{2}{6})^3 = 2^\frac{9}{6} + 3 \cdot 2^\frac{6}{6} \cdot 2^\frac{2}{6} + 3 \cdot 2^\frac{3}{6} \cdot 2^\frac{4}{6} + 2^\frac{6}{6} = \\ 2 \cdot 2^\frac{0}{6} + 6 \cdot 2^\frac{1}{6} + 6 \cdot 2^\frac{2}{6} + 2 \cdot 2^\frac{3}{6} $
  • $(2^\frac{3}{6}+2^\frac{2}{6})^4 = 2^\frac{12}{6} + 4 \cdot 2^\frac{9}{6} \cdot 2^\frac{2}{6} + 6 \cdot 2^\frac{6}{6} \cdot 2^\frac{4}{6} + 4 \cdot 2^\frac{3}{6} \cdot 2^\frac{6}{6} + 2^\frac{8}{6} = \\ 4 \cdot 2^\frac{0}{6} + 2 \cdot 2^\frac{2}{6} + 8 \cdot 2^\frac{3}{6} + 12 \cdot 2^\frac{4}{6} + 8 \cdot 2^\frac{5}{6} $
  • $(2^\frac{3}{6}+2^\frac{2}{6})^5 = 2^\frac{15}{6} + 5 \cdot 2^\frac{12}{6} \cdot 2^\frac{2}{6} + 10 \cdot 2^\frac{9}{6} \cdot 2^\frac{4}{6} + 10 \cdot 2^\frac{6}{6} \cdot 2^\frac{6}{6} + 5 \cdot 2^\frac{3}{6} \cdot 2^\frac{8}{6} + 2^\frac{10}{6} = \\ 40 \cdot 2^\frac{0}{6} + 40 \cdot 2^\frac{1}{6} + 20 \cdot 2^\frac{2}{6} + 4\cdot2^\frac{3}{6} + 2 \cdot 2^\frac{4}{6} + 10 \cdot 2^\frac{5}{6}$
  • $(2^\frac{3}{6}+2^\frac{2}{6})^6 = 2^\frac{18}{6} + 6 \cdot 2^\frac{15}{6} \cdot 2^\frac{2}{6} + 15 \cdot 2^\frac{12}{6} \cdot 2^\frac{4}{6} + 20 \cdot 2^\frac{9}{6} \cdot 2^\frac{6}{6} + 15 \cdot 2^\frac{6}{6} \cdot 2^\frac{8}{6} + 6 \cdot 2^\frac{3}{6} \cdot 2^\frac{10}{6} + 2^\frac{12}{6} = \\ 12 \cdot 2^\frac{0}{6} + 24 \cdot 2^\frac{1}{6} + 60 \cdot 2^\frac{2}{6} + 80 \cdot 2^\frac{3}{6} + 60 \cdot 2^\frac{4}{6} + 24 \cdot 2^\frac{5}{6} $

I am at a loss as to how this can help towards the solution... The first question in problem 18 is asking to prove the "rational root theorem" but I don't see how I can combine that with this hint to solve the problem.

UPDATE: following Gerry Myerson's advice I create a polynomial $c_0 \cdot 2^\frac{0}{6} + c_1 \cdot 2^\frac{1}{6} + c_2 \cdot 2^\frac{2}{6} + c_3 \cdot 2^\frac{3}{6} + c_4 \cdot 2^\frac{4}{6} + c_5 \cdot 2^\frac{5}{6}$ where each $c_i$ multiplies the coeffficients of the respective powers I found above for $x^0, \dots, x^6$.

So each upper-case $C_i$ is the sum of the coefficients from each of the powers $x_n$. For example $C_5 = 24 + 10c_5 + 8c_4 + 2c_2$. The expansion of this polynomial is:

$ \begin{aligned} C_0 \cdot 2^\frac{0}{6} + C_1 \cdot 2^\frac{1}{6} + C_2 \cdot 2^\frac{2}{6} + C_3 \cdot 2^\frac{3}{6} + C_4 \cdot 2^\frac{4}{6} + C_5 \cdot 2^\frac{5}{6} &&= \\ [ 12 + 40 \cdot c_5 + 4 \cdot c_4 + 2 \cdot c_3 + 2 \cdot c_2 + c_0] \cdot 2^\frac{0}{6} &+ &\\ [ 24 + 40 \cdot c_5 + 6 \cdot c_3] \cdot 2^\frac{1}{6} &+ & \\ [ 60 + 20 \cdot c_5 + 2 \cdot c_4 + 6 \cdot c_3 + 1 \cdot c_1] \cdot 2^\frac{2}{6} & + & \\ [ 80 + 4 \cdot c_5 + 8 \cdot c_4 + 2 \cdot c_3 + 1\cdot c_1 ] \cdot 2^\frac{3}{6}& + & \\ [ 60 + 2 \cdot c_5 + 12 \cdot c_4 + 1 \cdot c_2 ] \cdot 2^\frac{4}{6} & + & \\ [ 24 + 10 \cdot c_5 + 8 \cdot c_4 + 2 \cdot c_2 ] \cdot 2^\frac{5}{6} & & \\ \end{aligned} $

So as Danny Pak - Keung Chan explained, in order to use the rational root theorem I need to set each of these $C_i$ equal to zero and solve the system of equations:

$ \begin{aligned} 12 + 40 \cdot c_5 + 4 \cdot c_4 + 2 \cdot c_3 + 2 \cdot c_2 + c_0 = 0&\\ 24 + 40 \cdot c_5 + 6 \cdot c_3 = 0 \\ 60 + 20 \cdot c_5 + 2 \cdot c_4 + 6 \cdot c_3 + 1 \cdot c_1 = 0 \\ 80 + 4 \cdot c_5 + 8 \cdot c_4 + 2 \cdot c_3 + 1\cdot c_1 = 0 \\ 60 + 2 \cdot c_5 + 12 \cdot c_4 + 1 \cdot c_2 = 0\\ 24 + 10 \cdot c_5 + 8 \cdot c_4 + 2 \cdot c_2 = 0 \\ \end{aligned} $

Now this will hopefully yield integer solutions which I can replace into the polynomial and thus apply the rational root theorem (knowing that the value $\sqrt{2} + \sqrt[3]{2}$ is a root) which means it is either an integer or it is irrational.

At that point I guess I will just need to prove (easily verifying inequalities) that $\sqrt{2} + \sqrt[3]{2}$ is not an integer therefore it must be irrational.

This is a LONG way to solve this problem.


Update 2

I had GNU Octave solve the equations and got:

$c_0=-4. c_1=-24, c_2=12, c_3=-4, c_4=-6, c_5=0$

So our polynomial is:

$x^6 + 0x^5 -6x^4 -4x^3 +12x^2 -24x^1 -4$

This really does match the accepted answer's "minimal polynomial"

So now as I mentioned one can use part (a) of the problem (integral root theorem) to claim that the roots are either integral or irrational.

At this point it is easy to check that:

$1.4 < \sqrt{2} < 1.5$ and $1.2 < \sqrt[3]{2} < 1.3$ so adding we have:

$2.6 < \sqrt{2} + \sqrt[3]{2} < 2.8$

Therefore $\sqrt{2} + \sqrt[3]{2}$ is not an integer so by (a) it must be irrational and the long proof has come to its end.

  • 1
    The hints is poisonous. You do not need to expand up to power six. Don't follow the hints in the textbook. – Danny Pak-Keung Chan Aug 05 '20 at 01:38
  • 1
    The point of working out the powers is that you can use them to find integer coefficients $c_1,\dots,c_6$ such that $(\sqrt2+\root3\of2)^6+c_1(\sqrt2+\root3\of2)^5+\cdots+c_6=0$. Then you can apply the rational root theorem to this polynomial, $x^6+c_1x^5+\cdots+c_6$. – Gerry Myerson Aug 05 '20 at 01:39
  • To solve this question, we do not need any theorem. We just need the fact that $\sqrt2$ is irrational. The question can be solved with just few lines. – Danny Pak-Keung Chan Aug 05 '20 at 01:41
  • @GerryMyerson Can you expand on this a bit? How would I go about finding these coefficients using these powers? – Alexandros Aug 05 '20 at 02:12
  • @Alexandros The hint given by the textbook should appear in a textbook about Galois theory (exercises about field extension of $\mathbb{Q}$ by inserting surds). I don't know why it appears in a calculus textbook. Anyway, to work out the coefficients, you need to solve system of linear equations. – Danny Pak-Keung Chan Aug 05 '20 at 02:38
  • 1
    Take the equation I wrote, Alexandros, and plug in the expressions you got for the powers of $\sqrt2+\root3\of2$, multiply everything out, collect like terms to get something of the form $C_0+C_12^{1/6}+\cdots+C_52^{5/6}=0$, where each $C_i$ is an integer coefficient linear combination of the $c_i$. Set the $C_i$ all to zero to get a system of six linear equations in the six unknown $c_i$. Try it! – Gerry Myerson Aug 05 '20 at 06:07
  • @DannyPak-KeungChan Thanks I just got round to that by following Gerry Myerson's advice. I see that this technique is generic and could work with other numbers, so I find it quite interesting. I will update the question with the system of equations I've come up with. – Alexandros Aug 05 '20 at 17:31
  • @Alexandros That technique will appear again when you learn Galois theory and field extension. By the way, I don't know why it appears in a Calculus textbook. – Danny Pak-Keung Chan Aug 05 '20 at 17:36
  • @DannyPak-KeungChan Thank you for these pointers. I am now interested to learn more about Galois theory. – Alexandros Aug 05 '20 at 17:46
  • @GerryMyerson Can you please explain why $C_i$ has to be zero? – Orlin Aurum Oct 31 '22 at 05:25

2 Answers2

11

Let $a=\sqrt{2}+2^{\frac{1}{3}}.$ Prove by contradiction. Suppose the contrary that $a$ is rational. Observe that \begin{eqnarray*} 2 & = & (2^{\frac{1}{3}})^{3}\\ & = & (a-\sqrt{2})^{3}\\ & = & a^{3}-3a^{2}\sqrt{2}+3a\cdot2-2^{\frac{3}{2}}\\ & = & a^{3}+6a-\sqrt{2}(3a^{2}+2). \end{eqnarray*} Therefore, \begin{eqnarray*} \sqrt{2} & = & \frac{a^{3}+6a-2}{3a^{2}+2}\in\mathbb{Q}. \end{eqnarray*} It is well-known that $\sqrt{2}$ is irrational (Do I need to prove this fact?). Hence, we arrive a contradiction.

  • 2
    Very sharp answer... – imranfat Aug 05 '20 at 02:10
  • While this is correct, as I explained I am after the solution hinted toward in the textbook. This is a perfectly valid solution but does not use the hint. – Alexandros Aug 05 '20 at 02:13
  • 2
    @ Alexandros If you insist to follow the hint of the textbook and want to find a polynomial $p(x)$ with integer coefficients such that $p(a)=0$, you may write $\sqrt{2}(3a^2+2) = a^3 +6a-2$ and then take square on both sides. – Danny Pak-Keung Chan Aug 05 '20 at 02:52
1

Hint: I guess the idea is to Try to write $(\sqrt{2}+\sqrt[3]{2})^6$ in terms of $\sqrt{2}+\sqrt[3]{2}$ then we get a something like $$(\sqrt{2}+\sqrt[3]{2})^6 = p(\sqrt{2}+\sqrt[3]{2})$$ where $p(x)$ is a polynomial then using the rational zeros theorem we can show that $p(x)-x^6$ has no rational roots but $p(\sqrt{2}+\sqrt[3]{2})-(\sqrt{2}+\sqrt[3]{2})^6=0$. Thus we get that $\sqrt{2}+\sqrt[3]{2}$ is irrational.

To check your work see the link: https://www.wolframalpha.com/input/?i=minimal+polynomial+2%5E%281%2F2%29+%2B+2%5E%281%2F3%29

Since you still looking for the answer and you want to use your work I will start to write what you did using $b=\sqrt[6]{2}$ and $a=\sqrt{2}+\sqrt[3]{2}$ the your work

$$a^2=2+2b^5+b^4$$ $$a^3=2+6b+6b^2+2b^3$$ And so on the eliminates the $b$ powers do you want me to continue?

IrbidMath
  • 3,187
  • 12
  • 27
  • Such approach should appear in questions related to Galois theory. Why does a calculus textbook suggest it? – Danny Pak-Keung Chan Aug 05 '20 at 01:56
  • Rational zeros theorem is within the book I guess – IrbidMath Aug 05 '20 at 03:09
  • I don't know Galois theory but the book does ask to prove the specialization of the rational roots theorem in the same problem, in part (a) where the highest power coefficient is one (so the rational roots theorem degrades into an 'integral' roots theorem). I believe the exercise is structured as such because it is needed together with the hint in order to be solved in a specific way. – Alexandros Aug 05 '20 at 11:23
  • @AmerYR Your answer was too condensed for me to understand. But after DannyPak-KeungChan and Gerry Myerson walked me through it I now understand that you effectively pointed me in the correct direction. I've edited my question to spell out these huge polynomials and will accept your answer as correct. – Alexandros Aug 05 '20 at 17:49