# can you find the square root of n using only +, -, :, x ?

by logics
Tags: root, square
 P: 137 Is it possible to find √n using only the times table and the 4 operations?
 Mentor P: 4,164 Using only finitely many such operations, you cannot. For example taking the square root of two, after any such calculations you will be stuck with a rational number when you wanted an irrational one. You can calculate very good approximations though
P: 137
 Quote by Office_Shredder Using only finitely many such operations, you cannot....
It is possible:

if n = x² , digits up to 6, : you can find (in most cases) the exact result using no tools,
[likewise, you can find in all cases : $\sqrt[3]{x³}$ (9-digit n), and $\sqrt[5]{x^5}$ (15-digit n)]
if n = x² , digits > 6, : you can still guess three digits and get x with a few operations,

with a pocket calculator you can get any root in a few seconds. Can you believe that?

P: 3,015

## can you find the square root of n using only +, -, :, x ?

You can find the square root of a number a as the limit point of the following iterative procedure:
$$x_{n + 1} = \frac{1}{2} \, \left( x_n+ \frac{a}{x_n} \right)$$
However, as this is a limiting process, you can only get an exact result in the infinite limit $n \rightarrow \infty$. The convergence is fast though, and you get a good approximate result in just several iterations. Actually, you will exhaust the number of digits on your calculator pretty fast.
 P: 137 That is the Babylonian method, according to wiki. Is there a general algorithm that does not use powers of the same order (k) of n = x^k ? Cam you find$\sqrt{403225}$ using only logics, knowing that n= x²?
 P: 390 Why didn't you guys mention Taylor series? It's very easy to calculate the square-root with a series: (1+x)^n = 1 + n x + n(n-1) x^2/2! + n(n-1)(n-2) x^3/3! + ... Substitute n with 1/2 and there you got the square root. Keep adding terms till you get the precision that satisfied you. That's actually how computers calculate square-roots. http://en.wikipedia.org/wiki/Taylor_series
PF Patron
Emeritus
P: 16,094
 Quote by logics Is it possible to find √n using only the times table and the 4 operations?
It depends on what you mean by "find".

For example, for many purposes, an algorithm to produce arbitrarily good approximations (e.g. the method in this post) counts as "finding" the number.
PF Patron
Emeritus
P: 16,094
 Quote by TheDestroyer That's actually how computers calculate square-roots.
Reference? I find this claim rather unlikely....
Mentor
P: 4,164
 That's actually how computers calculate square-roots.
Are you sure about that? Taylor series are typically not a very fast way of calculating the function they are approximating. And your Taylor series only converges for |x|<1 which is pretty limited (the wikipedia article on 'calculating square roots' mentions how to adapt for this but also says the method is fairly slow compared to the others mentioned)

logics, if you just want the square root of something you know is an integer you can just use a binary search to find the square root without doing anything but multiplication a lot of times. For example to find the square root of 403225:

5002 = 250000 so we need a bigger number
7502 = 562500 so we need a smaller number. We try one in between these two:

6252 = 390625 so we need a number between 625 and 750
6872 = 471969

We need a number between 687 and 625
6502 = 422500

We need a number between 625 and 650:
6372 = 405769
We need a number between 625 and 635
etc, we find the square root is 635.

This technique can also be used to find the square root of any number to whatever precision you want, for example if you want the square root of 403226, we know that it is larger than 635 and smaller than 636. So we try 635.5, and if the square is too large, 635.25, and if the square is too large 635.125, and eventually when the number squared is smaller than 403226 we have a bound for the square root (and at that point we can keep dividing the interval in half in order to get more precision)
 PF Patron Sci Advisor Emeritus P: 16,094 Is it cheating to use the operation "<"?
P: 3,015
 Quote by Hurkyl Is it cheating to use the operation "<"?
What operation is that? What is the output of such an operation?
 Mentor P: 4,164 It's a binary operation, which takes two inputs a
P: 3,015
 Quote by Office_Shredder It's a binary operation, which takes two inputs a
your example is the midpoint method for solving the equation $x^2 - a = 0$. The Babylonian method is nothing but the Newton-Raphson method for solving the same equation.
P: 137
 Quote by Office_Shredder logics, if you just want the square root of something you know is an integer you can just use a binary search.....
When I say can you find x =$\sqrt{635²}$ using only logics I mean using only your mind, knowing by heart the times table, using no calculator, no pencil and paper.
HW Helper
P: 3,322
 Quote by logics When I say can you find x =$\sqrt{635²}$ using only logics I mean using only your mind, knowing by heart the times table, using no calculator, no pencil and paper.
I guess...

$635^2=403,225$ just for reference.

So we're trying to find the answer of $\sqrt{403,225}$. We know that since $100^2<403,225<1000^2$ then the answer will be a 3-digit number that we will denote as ABC.

To find the hundreds value, A, just realize that the first 2 digits of 403,225 are 40, and that is between 62 and 72 so A=6.

For C, the last 2 digits are 25, so I'd guess that it would be 5.

To find B - arguably the trickiest of all - I guess I'd flex my brain muscles by not pulling out a pen and paper and expanding (600+10B+5)2 in my mind, and scrapping all the values except the relevant ones that would be the biggest influence on the ten-thousand's value in the answer.

So $$(600+10B+5)^2=(600+(10B+5))^2=600^2+1,200(10B+5)+(10B+5)^2$$

Clearly the influence would be that middle term $1,200(10B+5)$

So now let's do some simplifications. $1,200(10B+5)>12,000B$ and $403,225-360,000 \approx 40,000$ and then $$3<\frac{40,000}{12,000}<4$$ and so I would make the logical guess that B=3.

Thus, ABC = 635
P: 137
 Quote by Mentallic So $$(600+10B+5)^2=(600+(10B+5))^2=600^2+1,200(10B+5)+(10B+5)^2$$ Clearly the influence would be that middle term $1,200(10B+5)$......Thus, ABC = 635
Well done, Mentallic, if you can manage all that in your mind!
But you must find something simpler if you want to solve the challenge in the twin-thread about the cube root... (try to find x = $\sqrt[3]{723³}$).
If you do not wish to move on to that thread, here is a slightly tougher [but still easy] challenge: find x = $\sqrt{277729}$
HW Helper
P: 3,322
 Quote by logics Well done, Mentallic, if you can manage all that in your mind! But you must find something simpler if you want to solve the challenge in the twin-thread about the cube root... (try to find x = $\sqrt[3]{723³}$). If you do not wish to move on to that thread, here is a slightly tougher [but still easy] challenge: find x = $\sqrt{277729}$
With a bit of practice, I could answer the square root problems more efficiently, as long as they're either 4 or 6 digits, which is pretty narrowed down and a useless skill to have if you ask me.

And obviously I'd have little luck extending it to the cube roots.
 Sci Advisor P: 1,670 Do you "find" sqrt(2) by the operation sqrt?

 Related Discussions Calculus & Beyond Homework 5 Introductory Physics Homework 6 General Math 9 Introductory Physics Homework 14 Introductory Physics Homework 10