9

Solve the equation $$x+\sqrt{a+\sqrt{x}}=a$$ for all real $x$ and nonzero parameter $a$, prove that the equation has a solution if and only if $a\ge 1$.

My attempts and thoughts.

$$\sqrt {a+\sqrt x}=a-x$$

$$a+\sqrt x=(a-x)^2$$

$$\sqrt x =(a-x)^2-a$$

$$x=\big((a-x)^2-a\big)^2$$

$$x=(a^2-2ax+x^2-a)^2$$

But the last expansion seems terrible.

I haven't been able to prove why $a\ge 1$ has to be. My other concern is that not all roots of the last equation may be valid.For example, say $\sqrt {x}=-1$. We have $x=1$ but $x=1$ doesn't satisfy $\sqrt {x}=-1$.

My questions. Is there a method that doesn't square the equation and works easily in general? How can we get rid of excess roots? How can we prove that $a\ge 1$ ?

Bill Dubuque
  • 282,220
  • 12
    Does this answer your question? Solving $x+\sqrt{a+\sqrt{x}}=a$ for $x$. - found using an Approach0 search. Note there's also the AoPS thread Solve the following equations for real 'x'. – John Omielan May 24 '23 at 21:04
  • @JohnOmielan But $a\ge 1$ is a different question and I am looking for non square methods –  May 24 '23 at 21:13
  • 1
    If $x=0$ the left hand is $\sqrt a$ and it clearly grows to infinity as $x$ grows. Thus there is a solution if and only if $\sqrt a≤a$. – lulu May 24 '23 at 21:19
  • 2
    Note: not sure I agree with the close votes. The supposed duplicate is really answering a different question (looking for a closed form) but the OP has asked for an existence result which one might (accurately) expect to be a lot simpler. – lulu May 24 '23 at 21:21
  • Yes @lulu the duplicate version does not answer my question please open my question –  May 24 '23 at 21:26
  • 1
    @user1178051 I mistakenly thought the answer in the proposed duplicate included what you're looking for. However, on looking at it again more carefully, it doesn't really do that. Also, as lulu's comment indicates, there's a relatively simple solution for your problem (that the comment itself actually included). As such, I apologize and have voted to reopen your question. – John Omielan May 24 '23 at 21:30
  • 3
    Just voted to reopen. – lulu May 24 '23 at 21:34
  • In the linked previous solution we take $x=(y^2-a)^2$ for some $y$, we are not squaring the equation. Why isn’t it a good solution for your question? – user May 25 '23 at 07:04
  • 1
    @JohnOmielan: That comment solves the “prove that the equation has a solution if and only if $a \ge 1$” part, but as I understand it, the question also asks how to determine the solution. – Martin R May 25 '23 at 07:35
  • https://math.stackexchange.com/questions/4703549/solve-for-x-x-sqrta-sqrtx-a-where-a1/4703553#4703553 – nonuser Jun 01 '23 at 05:27

10 Answers10

5

$$x+\sqrt{a+\sqrt{x}}=a$$

Move term and square it, we get

$$a+\sqrt x=(a-x)^2$$

The trick is we can treat this as a quadratic equation with respect to $a$ $$a^2-(2x+1)a+x^2-\sqrt x=0$$ Instead of solving $x$, we solve for $a$

$$a=\frac{2x+1\pm\sqrt{4x+4\sqrt x+1}}{2}$$ Note that inside square root is a complete square $(2\sqrt x+1)^2$, so we get $$a=\frac{2x+1\pm(2\sqrt x+1)}{2}$$

Let $t=\sqrt x$ and we get two equations:

$$\begin{cases} \color{blue}{a=t^2+t+1}\\ \\ \color{red}{a=t^2-t} \end{cases}$$

First let's look at the $\color{red}{\text{red-color}}$ equation. We will show it has no real roots.

$$x+\sqrt{a+\sqrt{x}}=a$$

Here, the OP implies $$\sqrt{a+\sqrt{x}}=a-x\ge0\Rightarrow a\ge x\Rightarrow \color{green}{a\ge t^2}\tag{1}$$

Plug in the $\color{red}{\text{red-color}}$ equation, we get

$$\color{red}{t^2-t=a}\ge t^2\Rightarrow -t\ge 0\Rightarrow -\sqrt x\ge0$$

This is true if and only if $x=0$, so $t=0$, the $\color{red}{\text{red-color}}$ equation gives $a=0^2+0=0$

But we require $a\neq 0$, hence we get a contradiction. Therefore, the $\color{red}{\text{red-color}}$ equation has no real roots. We can ignore it.

Now let's move to the $\color{blue}{\text{blue-color}}$ equation. Note the restriction "$\color{green}{a\ge t^2}$" in eq.(1) is automatically satisfied here.

$$\color{blue}{a=t^2+t+1}$$

Solve it and we get:

$$t=\frac{-1\pm\sqrt{4a-3}}{2}$$

Since $t=\sqrt x\ge0$, we only keep the $+$ root.

$$t=\frac{-1+\sqrt{4a-3}}{2}$$

Finally, to determine parameter $a$, we need following conditions:

$$4a-3\ge0~~\cap~~t\ge0$$

This gives the range for $a$

$$\boxed{a\ge1}$$

The solution is:

$$\boxed{x=t^2=\frac{2a-1-\sqrt{4a-3}}{2}}$$

MathFail
  • 21,529
4

Let us know in the comments whether this answer solves your concerns and the answer can be improved later .


Since the left-hand side of the original equation is always non-negative, we observe that $a>0$ . Then, we want to add $\sqrt x$ to both side, to check if it works for us :

$$ \begin{align}x+\sqrt x+\sqrt{a+\sqrt{x}}&=a+\sqrt x\end{align} $$

Substituting $\thinspace y:=\sqrt x$ and $z:=\sqrt {a+\sqrt x}=a-y^2$, where $y≥0\wedge z>0\thinspace $ then you have :

$$ \begin{align}&{\color{#c00}{y^2}}+{\color{#0a0}{y}}+{\color{#0a0}{z}}-{\color{#c00}{z^2}}=0\\ \implies &(y+z)(y-z+1)=0\\ \implies &y=z-1=a-y^2-1\end{align} $$

Since $\thinspace a>0$ and $y≥0$, then you get the following :

$$ \begin{align}&y^2+y-(a-1)=0\\ \implies &y=\frac {-1+\sqrt {4a-3}}{2}\\ \implies &x=\frac {\left(\sqrt {4a-3}-1\right)^2}{4}\end{align} $$

Thus, the final result becomes :

$$\bbox[5px,border:2px solid #C0A000]{x=\frac {2a-\sqrt {4a-3}-1}{2}}$$

This implies that, you have only one real root ( solution ) .

Finally, to prove that $a≥1$, observe :

$$ \begin{align}y≥0&\iff\sqrt {4a-3}≥1\\ &\iff a≥1\end{align} $$

which completes the answer .

  • 1
    At $y^2+y-(a-1)=0$ you can already conclude that $a = y^2+y+1 \ge 1$. In particular, $\sqrt{4a-3}$ is defined and the subsequent calculations are justified. That is (IMO) better than “just computing” the solution and showing $a \ge 1$ in retrospect. – Martin R May 25 '23 at 08:21
  • 1
    I had $2$ reasons here. The OP used the statement "if and only if" instead of "if" in the question. So, showing that "Since $y≥0$ , then $a=\dots +1≥1$" was not sufficient . The second justification was : the original question statement was stating : "... the equation has a solution, if and only if, when $a≥1$ ". Therefore, I planned to show this fact after exactly providing the non-negative closed-form solution for $y$ . – lone student May 25 '23 at 19:54
  • @lone student, if you are going to be pedantic like that, you should say that these implications are equivalences since you are not going to just plug in that expression for $x$ in the starting equation to confirm that it is a solution. – Ennar May 27 '23 at 13:01
  • Having that in mind and what Martin R said, you have that a solution exists if and only if $a = 1 + y + y^2,\ y \geq 0,\ a > 0$, which is obviously equivalent to $a = 1 + y + y^2,\ y \geq 0,\ a \geq 1$. – Ennar May 27 '23 at 13:15
  • 1
    @Ennar Well, but how do you know that, $y^2+y+1=a$ has a solution ? – lone student May 27 '23 at 13:20
  • Well, you wrote that part, I was just referring to what Martin said. Concretely, $a = 1 + y + y^2,\ y\geq 0,\ a\geq 1$ is equivalent to $y = \frac {-1+\sqrt {4a-3}}{2}, \ a \geq 1$. So, now you do know that $a\geq 1$ is not just necessary, it's also sufficient condition for a solution to exist, since you gave explicit formula. I'm not saying I'm adding anything new to what you wrote, I'm just giving a suggestion how you can make your answer more tidy (imo), since OP started a bounty, and I think your answer is canonical and deserves the bounty. – Ennar May 27 '23 at 13:28
4

I'll start with the second part first. If you're happy with graphical "proofs", you could consider the two graphs $y=a-x$ and $y=\sqrt{a+\sqrt{x}}$. You could probably make this more formal using the intermediate value theorem and other calculus arguments.

Case 1: $a>1$. Here you have $a>\sqrt{a}$, so you get exactly one intersection point where $x>0$.

enter image description here

Case 2: $a=1$. Here you have $a=\sqrt{a}$, so the two graphs intersect at $x=0$.

enter image description here

Case 3: $0<a<1$. Here you have $a<\sqrt{a}$, so since $y=\sqrt{a+\sqrt{x}}$ is increasing and $y=a-x$ is decreasing, there is no intersection point.

enter image description here

Case 4: $a<0$. Lastly, it's clear that when $a<0$ that there are no solutions, since we have $x\ge 0$ from the natural domain of the square root function, so the left-hand side $x+ \sqrt{a+\sqrt{x}}$ is non-negative, while the right-hand side is $a$ is negative.

The combination of the four cases shows that the equation has a solution if and only if $a\ge 1$.

For your first part, I don't have an solution which eliminates the 'squaring problem', but here's one that avoids the 'quartic problem'. One possible trick here to save yourself from solving a quartic is to solve it for $a$ rather than $x$. So if we go ahead and move the $x$ to one side and square it:

$$ a+ \sqrt{x} = (a-x)^2 \Rightarrow a^2 - (2x+1) a + (x^2 - \sqrt{x}) = 0.$$

Using the quadratic formula, you'll end up getting $$ a = x + \sqrt{x} + 1, x - \sqrt{x}.$$

Because we know that $a=1, x=0$ is a solution, we can eliminate the second option, so now we have a quadratic equation in $\sqrt{x}$, namely $(\sqrt{x})^2 + \sqrt{x} + (1-a) = 0$. Solving this, we get $$\sqrt{x} = \frac{-1\pm \sqrt{4a-3}}{2} \Rightarrow x = \frac{2a-1 \pm \sqrt{4a-3}}{2}.$$

Again, knowing that $a=1, x=0$ is a solution shows us that $x = \dfrac{2a-1 + \sqrt{4a-3}}{2}$.

Gary Liang
  • 1,079
2

This is another form of graphical method. This answer proves $a \ge 1$ is required for solutions only, as a few other answers already target obtaining mathematical solutions to your equation. This answer may not be something you can use mathematically, but it fulfils your wish of:

I want a simpler and more descriptive answer to my concerns

Consider the LHS and RHS as separate functions, ie.

$y = a \\ y = x + \sqrt{a + \sqrt{x}}$

Looking at it this way, we need to find when the line $y = a$ has intersection(s) with the function $\sqrt{a + \sqrt{x}}$. This would be the same as when both graphs take the same value, and hence you obtain your solution for $a = x + \sqrt{a + \sqrt{x}}$.

$y = a$ is not much of an interesting function, it remains constant over $x \in [0, \infty)$. However, the function $y = x + \sqrt{a + \sqrt{x}}$ is more intriguing to analyse. Let $f(x) = x + \sqrt{a + \sqrt{x}}$.

Note that $f(x)$ only exists for $x \ge 0$.

Now see, $$\large f(x)= \underbrace{x}_{\color\green+} + \underbrace{\sqrt{a + \sqrt{x}}}_{\color\green+}$$

and hence, $\displaystyle{\color\red{f(x) > 0}}$ regardless of $a$ and strictly increasing. This suggests the $\min{(f(x))}$ occurs at $\displaystyle{\color\red{x = 0}}$

Given $f(0) = \sqrt{a}$, see that an intersection would occur if

$y = a = \sqrt{a} = f(x = 0)$

or

$y = a > \sqrt{a}$ as $f(x)$ would still exist such that $y = a = f(x)$ has a solution/intersection for some $+x$. Observe, how the info that $f(x)$ is strictly increasing plays a role here.

Overall and finally, we look for such $a$ satisfying $a \ge \sqrt{a}$; only to occur when $$\large{a \ge 1}$$

Dstarred
  • 3,065
  • 7
  • 26
1

Reading the answers, I first thought that some answers show an incorrect proof for the claim that the equation has a solution if and only if $a\ge 1$.

However, "incorrect" is perhaps too strong a word.

I'm going to write a solution first, and then show some thoughts about a solution, and finally write another solution which tries to avoid using "squaring".


A solution :

From $$x+\sqrt{a+\sqrt x}=a\tag1$$ we have to have $$x\ge 0\qquad\text{and}\qquad a\gt 0\tag2$$ ($a\gt 0$ because LHS $\ge 0$, and $a$ is non-zero).

Under $(2)$, we have (as some other answers did) $$\begin{align}(1)&\iff \sqrt{a+\sqrt x}=a-x \\\\&\iff a+\sqrt x=(a-x)^2\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a^2+(-2x-1)a+x^2-\sqrt x=0\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a=\frac{2x+1\pm\sqrt{4x+4\sqrt x+1}}{2}\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a=\frac{2x+1\pm\sqrt{(2\sqrt x+1)^2}}{2}\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a=\frac{2x+1\pm (2\sqrt x+1)}{2}\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a=x+\sqrt x+1,x-\sqrt x\qquad\text{and}\qquad a-x\gt 0 \\\\&\iff a=x+\sqrt x+1\tag3\end{align}$$

Now, let us prove that $(1)$ has a solution if and only if $a\ge 1$.

Since we've already seen that $(1)$ has a solution if and only if $(3)$ has a solution, it is sufficient to prove that $(3)$ has a solution if and only if $a\ge 1$.

  • If $(3)$ has a solution, then $a=x+\sqrt x+1\ge 1$.

  • If $a\ge 1$, then $$\begin{align}(3)&\iff \bigg(\sqrt x+\frac 12\bigg)^2=a-1+\frac 14\\\\&\iff \underbrace{\sqrt x+\frac 12}_{\color{red}{\text{positive}}}=\color{red}+\sqrt{\underbrace{a-1}_{\ge\ 0}+\frac 14}\ \ \bigg(\ge\sqrt{\frac 14}=\frac 12\bigg)\end{align}$$ so $(3)$ has a solution.

It follows that $(3)$ has a solution if and only if $a\ge 1$.

Therefore, we can say that $(1)$ has a solution if and only if $a\ge 1$.

Since we get $\sqrt x=\sqrt{a-\frac 34}-\frac 12$, we finally obtain $$x=\bigg(\sqrt{a-\frac 34}-\frac 12\bigg)^2=\frac{2a-1-\sqrt{4a-3}}{2}$$


Next, I'm going to show some thoughts about the following solution.

"We have $x+\sqrt x+1=a$. Solving this, we get $\sqrt x=\frac{-1\pm\sqrt{4a-3}}{2}$. Since we can exclude $\frac{-1-\sqrt{4a-3}}{2}$, we get $\sqrt x=\frac{-1+\sqrt{4a-3}}{2}$. A necessary and sufficient condition for this to be real and non-negative is $4a-3\ge 0$ and $\frac{-1+\sqrt{4a-3}}{2}\ge 0$, i.e. $a\ge 1$."

I first thought that this solution is incorrect. The reason is as follows :

$\sqrt x=\frac{-1+\sqrt{4a-3}}{2}$ is valid only when $a\ge 1$ (since LHS is real and non-negative). So, when one writes $\sqrt x=\frac{-1+\sqrt{4a-3}}{2}$, one implicitly assume $a\ge 1$ (If $a\lt 1$, then $\sqrt x=\frac{-1+\sqrt{4a-3}}{2}$ does not make sense).

Then, I received the following comment :

It looks to me that you are ignoring the fact that quadratic equation always has two complex solutions. One can observe a posteriori that out of these two solutions, there is a real nonnegative solution if and only if $a≥1$.

It seems that this comment suggests the following :

"We have $x+\sqrt x+1=a$. There are two complex solutions $\sqrt x=\frac{-1\pm\sqrt{4a-3}}{2}$. Since we can exclude $\frac{-1-\sqrt{4a-3}}{2}$, we get $\sqrt x=\frac{-1+\sqrt{4a-3}}{2}$. A necessary and sufficient condition for this to be real and non-negative is $4a-3\ge 0$ and $\frac{-1+\sqrt{4a-3}}{2}\ge 0$, i.e. $a\ge 1$."

This makes sense, so "incorrect" is perhaps too strong a word.

Some comments :

  • Since the "solution" in "the equation has a solution if and only if $a\ge 1$" is real, I thought that we have to solve our problem only in real numbers.

  • There is a short solution only in real numbers. For example, see the above solution of mine. So, our problem can be solved only in real numbers.

  • I not only prefer a solution only in real numbers especially for our problem, but also believe that we are implicitly asked to prove the claim only in real numbers, but I accept the fact that the problem does not say so.


A solution which tries to avoid using "squaring" :

Let $t:=\sqrt{a+\sqrt x}\gt 0$, so $t^2=a+\sqrt x$.

$$(1)\iff x+t=a$$

Eliminating $a$, we get $$x+t=t^2-\sqrt x\iff t^2-t-x-\sqrt x=0$$

Solving this for $t$ gives $$t=\frac{1\pm \sqrt{(2\sqrt x+1)^2}}{2}=\frac{1\pm (2\sqrt x+1)}{2}=1+\sqrt x,-\sqrt x$$ Since $t\gt 0$, we get $$(t=)\ \ a-x=1+\sqrt x\tag3$$

Note that if $(3)$ has a solution, then $(1)$ has a solution since $$\sqrt{a+\sqrt x}=\sqrt{1+2\sqrt x+x}=1+\sqrt x=a-x$$ The rest is the same as the solution written at the top.

mathlove
  • 151,597
  • 1
    It looks to me that you are ignoring the fact that quadratic equation always has two complex solutions. One can observe a posteriori that out of these two solutions, there is a real nonnegative solution if and only if $a\geq 1$. – Ennar May 28 '23 at 15:22
  • @Ennar : Thanks. Your comment is helpful. I thought that we have to solve our problem only in real numbers. I prefer a solution only in real numbers especially for our problem, though. – mathlove May 29 '23 at 08:12
  • 1
    No, we do need to solve this particular problem in nonnegative reals, but that doesn't mean that we can't use quadratic formula and decide under which conditions we have a nonnegative real solution. Let me give you a similar example just to be clear. Let's say I want to determine all integer solutions of the equation $2x^2-3x-2 = 0$. By the rational root theorem, the candidates for rational solutions are ${\pm1,\pm2,\pm 1/2}$. By direct inspection, only $2$ is an integer solution. Wanting integer solutions only doesn't make rational numbers not exist or rational root theorem invalid. – Ennar May 29 '23 at 12:42
1

Making $x = y^2$ we have

$$ (y^2-a)^2= y+a\Rightarrow y^4-2ay^2-y+a^2-a = (y^2-a-y)(y^2-a+y)+y^2-y-a =(y^2-y-a)(y^2+y-a+1) $$

now solving $y^2-y-a = 0$ we have $y = \frac{1}{2} \left(1\pm\sqrt{4 a+1}\right)\ge 0$ then solving $1\pm\sqrt{4 a+1}=s^2$ for $a$ we have

$$ a = \frac{1}{4} \left(s^2-2\right)s^2 $$

but $a$ should be non negative, so we rule out this possibility.

Taking now $y^2+y-a+1=0$ we have $y = \frac{1}{2} \left(-1\pm\sqrt{4 a-3}\right)$ then analogously we need $\left(-1\pm\sqrt{4 a-3}\right)=s^2$ that solved for $a$ gets

$$ a = \frac{1}{4} \left(4+(s^2 + 1)s^2\right) $$

or $a\ge 1$

Cesareo
  • 36,341
0

$$x = a - \sqrt{a +\sqrt x}$$

$$= a - \sqrt{a + \sqrt{a - \sqrt{a + \sqrt{x}}}} $$

=$ a - \sqrt{a + \sqrt{a - \sqrt{a + \sqrt{a ....}}}}$till infinity

let $y$= $\sqrt{a + \sqrt{a - \sqrt{a + \sqrt{a - ...}}}}$till infinity

$y^2 = a + \sqrt{a - \sqrt{a +.......}}$

$(y^2 - a )^2 = a - \sqrt{a + \sqrt{a...}}$

$$y^4 + a^2 -2ay^2 - a = -y$$

therefore $ y^4 - 2ay^2 + y + a^2 - a = 0$, $ y > 0$, $a - y > 0$, $a > 0$

But I'm sorry this is still an incomplete solution but I can't comment so I posted my idea. Hope it will help some other person to solve this. Sorry again:(

0

As already proposed in the answer for the linked question, we have that by $x=(y^2-a)^2$ with $y^2-a\ge 0$ and $y\ge 0$

$$x+\sqrt{a+\sqrt{x}}=a \iff ( y^2 + y-a) ( y^2 - y + 1-a) = 0$$

which lead respectevely to

  • $\color{red}{y=\frac{-1- \sqrt{4a+1}}{2}<0\;\text{(no solution)}}$
  • $\color{red}{y=\frac{-1+ \sqrt{4a+1}}{2}\ge 0 \;\text{with}\;a\ge 0\implies y^2-a=\frac{1- \sqrt{4a+1}}{2}<0\;\text{(no solution)}}$
  • $\color{red}{y=\frac{1- \sqrt{4a-3}}{2}\;\text{with}\;\frac34 \le a\le1\implies y^2-a=\frac{-1- \sqrt{4a-3}}{2}<0\;\text{(no solution)}}$

and finally

  • $\color{green}{y=\frac{1+ \sqrt{4a-3}}{2}>0\;\text{with}\;a\ge \frac34 \implies y^2-a=\frac{-1+ \sqrt{4a-3}}{2}\ge 0\;\text{with}\;a\ge 1}$

which leads to the unique solution

$$\boxed{x=(y^2-a)^2=a-\frac{1+\sqrt{4a-3}}{2}}$$

for $a\ge 1$.

user
  • 162,563
0

Let $\quad u=-\sqrt{a+\sqrt x}\le0,\quad v=\sqrt x\ge0,\quad$ then $$u^2-v=v^2-u=a,$$ $$\left(u+\frac12\right)^2 = \left(v+\frac12\right)^2=a+u+v+\frac14,\quad u\le 0,\quad v\ge 0,$$ $$u+\dfrac12=-\left(v+\frac12\right),\quad u+v=-1,$$ $$v^2+v+1-a=0,\quad v\ge0\Rightarrow\quad v=\sqrt{a-\frac34}-\frac12,\quad a\ge1,$$ $$\color{brown}{\mathbf{x=v^2=a-\frac12-\sqrt{a-\frac34},\quad a\ge1}.}$$

0

Here is a solution that comes (arguably) under the conditions of the question, namely "do not square", and while searching for solutions $x\ge 0$ of the given equation $$ f(x) = 0\ ,\qquad\ f:\Bbb R_{\ge 0}\to \Bbb R\ ,\ f(x) = x+\sqrt{a+\sqrt x} - a\ , $$ how to "eliminate excess roots", i.e. roots that would pop up after getting an equivalent polynomial equation in $x$ of higher degree after taking squares. Also, we want a formula for the solution. The idea is to use the Galois conjugated expressions, so we get some structure during the computations.


  • $f$ is continuous, strictly increasing, and $f(0)=\sqrt a-a$. The value $a=0$ was excluded. If $a\in(0,1)$, then $f(0)>0$, so there is no solution. Also, for $a=1$ we have only the solution $x=0$, $f$ being monotone.
  • So we do assume from now on $a> 1$. Then $f(0)<0<f(a)$, so there exists a unique solution in $(0,a)$.
  • Let $x$ be this solution for the equation $f=0$. We "do not square", but use the conjugate expressions $E_\pm\pm$, and $f=0$ is equivalent to $E_{-+}=0$, getting: $$ \begin{aligned} E_{++}(a,x) &= (a-x) + \sqrt{a+\sqrt x}\ ,\\ E_{-+}(a,x) &= (a-x) - \sqrt{a+\sqrt x}\ ,\\ E_{+-}(a,x) &= (a-x) + \sqrt{a-\sqrt x}\ ,\\ E_{--}(a,x) &= (a-x) - \sqrt{a-\sqrt x}\ ,\qquad\text{ and compute products:}\\[3mm] E_{++} E_{-+}\ (a,x) &= (a-x)^2 - (a+\sqrt x)\ ,\\ E_{+-} E_{--}\ (a,x) &= (a-x)^2 - (a-\sqrt x)\ ,\qquad\text{ and finally:}\\[3mm] \Pi E(a,x):= E_{++} E_{-+} E_{+-} E_{--}\ (a,x) &= \Big(\ (a-x)^2 - a\ \Big)^2 - x\ . \end{aligned} $$ Yes, equation to zero, this is also the true polynomial equation obtained when twice building squares, but there is a further plus in this "Galois-theoretic" view.

Observe that if we introduce correspondingly instead of $a$ values $A_\pm$ given by $$\bbox[yellow]{A_\pm=x\pm\sqrt x} $$ we obtain: $$ \begin{aligned} E_{\pm-}(A_+,x) &= (A_+-x) \pm \sqrt{A_+-\sqrt x}\\ &=((x+\sqrt x)-x)\pm \sqrt{(x+\sqrt x)-\sqrt x}=\sqrt x\pm\sqrt x\ ,\\ E_{\pm+}(A_-,x) &= (A_--x) \pm \sqrt{A_-+\sqrt x}\\ &=((x-\sqrt x)-x)\pm \sqrt{(x-\sqrt x)+\sqrt x}=-\sqrt x\pm\sqrt x\ ,\\ \end{aligned} $$ so two factors above are vanishing. I hope this observation is a fair one, this is the key point in the solution. Without this point, the solution is missing the point. The final product $\prod E_{\pm\pm}$ is a polynomial, so we expect in this product the polynomial factor $$ (a-A_+)(a-A_-)=a^2-(A_++A_-)a+A_+A_-=a^2-2xa+(x^2-x)=((a-x)^2-x)\ . $$ And of course, if we know this factor, we can force it: $$ \begin{aligned} \Pi E(a,x) &= \Big(\ (a-x)^2 - a\ \Big)^2 - x \\ &=(a-x)^4\color{gray}{-x(a-x)^2+x(a-x)^2}-2a(a-x)^2+a^2 - x \\ &=((a-x)^2-x)\Big((a-x)^2-(2a-x)\Big) \color{gray}{-x(2a-x)}+a^2 - x \\ &=((a-x)^2-x)\color{blue}{\Big((a-x)^2-(2a-x)+1\Big) } \ . \end{aligned} $$ So the given equation $f=E_{-+}=0$ is equivalent to the vanishing of the blue polynomial, $0= x^2-(2a-1)x+(a-1)^2$. Among the two solutions $$ \bbox[lightblue]{\frac 12\Big((2a-1)-\sqrt{4a-3}\Big)}\ ,\qquad \frac 12\Big((2a-1)+\sqrt{4a-3}\Big)>\frac 12\Big((2a-1)+\sqrt{4-3}\Big)=a\ , $$ we have to refuse the last one, since it lives outside $(0,a)$. So the one in the light blue color remains.

$\square$



Note: The exposition would have been shorter, if allowed to observe that plugging in instead of $a$ the values $$\bbox[yellow]{B_\pm=x\pm\sqrt x+1} $$ we obtain annihilation in two cases, and then consider only these cases: $$ \begin{aligned} E_{\pm+}(B_+,x) &= (B_+-x) \pm \sqrt{B_+ + \sqrt x}\\ &=((x+\sqrt x+1)-x)\pm \sqrt{(x+\sqrt x+1)+\sqrt x}=\sqrt x+1\pm\sqrt {(\sqrt x +1)^2}\ ,\\ E_{\pm-}(B_-,x) &= (B_--x) \pm \sqrt{B_- - \sqrt x}\\ &=((x-\sqrt x+1)-x)\pm \sqrt{(x-\sqrt x+1)-\sqrt x}=-\sqrt x+1\pm\sqrt {(\sqrt x -1)^2}\ . \end{aligned} $$

dan_fulea
  • 37,952