2

$$2^{x}+3^{x}=6$$

How to find the real number x? I mean it's approximately $1.19$ bur can we write $x$ as the form of $a, b, c$ when $a^{x}+b^{x}=c$ in general. Maybe an infinite sum?

esege
  • 3,621

2 Answers2

6

This kind of equation does not show analytical solution and only numerical methods would be able to solve the problem.

What is nice if that we know the range which contains the bounds $$1 < x < \frac{\log (3)}{\log (2)}$$

However, since the function is quite stiff (it varied very fast), it should be better to consider the function $$f(x)=\log(2^x+3^x)-\log(6)$$ which looks like a straight line. So, let me be lazy and start iterating from $x_0=1$. Then Newton iterates will be $$x_1=1.19469932382260$$ $$x_2=1.19391148996091$$ $$x_3=1.19391147721159$$ which is the solution for fifteen significant figures.

For the more general case of $a^x+b^x=c$, assuming $a>0,b>0,c>0$, we know that the solution will be such that $$\min \left(\frac{\log \left(\frac{c}{2}\right)}{\log (a)},\frac{\log \left(\frac{c}{2}\right)}{\log (b)}\right)<x<\max \left(\frac{\log \left(\frac{c}{2}\right)}{\log (a)},\frac{\log \left(\frac{c}{2}\right)}{\log (b)}\right)$$ and, starting at the middle of the range, we can start Newton procedure for $$f(x)=\log(a^x+b^x)-\log(c)$$ If we can start using $x_0=1$, the first iterate of Newton method will be $$x_1=\frac{-(a+b) (\log (a+b)-\log (c))+a \log (a)+b \log (b)}{a \log (a)+b \log (b)}$$

4

There is a way to build an (approximate) explicit formula, using Euler's method of substitution (I think Euler was the first to use it anyway).

First, we need to transform our equation. Let's take $3^x=y$ and $\alpha=\frac{log(2)}{log(3)}$ with $log$ being the natural logarithm. Then we have:

$$ y^\alpha+y-6=0 $$

$$ y=6-y^\alpha $$

$$ y^\alpha=(6-y^\alpha)^\alpha $$

Substitute $y^\alpha$ into the right side several times:

$$ y \approx 6-\left(6-\left(6-\left(6-\left(6-... \right)^\alpha \right)^\alpha \right)^\alpha \right)^\alpha $$

Then we calculate $\alpha=0.63092975357$ and calculate $y$.

This expression is very inefficient - it converges very slowly, but it will get you the approximate value easily enough. The accurate solution is:

$$ y=3^x=3^{1.1939115}=3.71228 $$

The first few values given by our expression are:

$$ y_1=6-6^\alpha=2.903... $$

$$ y_2=6-(6-6^\alpha)^a=4.041... $$

$$ y_3=3.586... $$

$$ y_4=3.762... $$

$$ y_5=3.693... $$

$$ y_6=3.720... $$

$$ y_7=3.709... $$

As you can see, it's pretty bad. But seeing as odd values are smaller than the accurate solution and even values are larger, you can use the mean value for better approximation.

Anyway, you wanted the general expression for $a^x+b^x=c$ - here is the way you can make it.

Please note, that for convergence the following condition should hold: either $c>1$ and $\alpha<1$ or $c<1$ and $\alpha>1$. For any other cases, the expression will not work.

Yuriy S
  • 31,474
  • I have no idea how you arrive at the equality right before you mention the value of $\alpha$. – Wojowu Feb 03 '16 at 13:01
  • @Wojowu, it's not supposed to be an equality, I edited the answer accordingly. Thanks – Yuriy S Feb 03 '16 at 13:03
  • In that case, I don't see at what point you justify your claim from the first sentence "There is a way to build explicit formula...". At least in my opinion, the limit you give is not explicit as it still involves a limit. – Wojowu Feb 03 '16 at 13:05
  • @Wojowu, it's an explicit formula because it allows to compute an approximate solution explicitly. Claude Leibovici offered a better method of solving the equation, I just offered a method which is easier to understand – Yuriy S Feb 03 '16 at 13:12
  • For me, the term "explicit formula" coincides with the term "closed form", so I disagree about it being an explicit formula for the number in OPs question. I do, however, agree that it gives us explicit formulas for some numerical approximations. – Wojowu Feb 03 '16 at 13:16
  • @Wojowu I edited the first sentence so there is no confusion. I would like to keep the term 'explicit', even at the price of some downvotes. Thank you anyway for your help – Yuriy S Feb 03 '16 at 13:19