Thursday, April 28, 2011

In Case You Wondered…

Very recently I discussed fractional exponents. In that routine I used only multiplications, divisions, and pulling square roots. But what about pulling square roots by hand? There is a simple formula for that. Alas, it is iterative. Here it is:

square root (eventually) = ½ * ((number /guess) + guess)

A straightforward “guess” is simply to divide the number by two. Let me demonstrate this formula by iteration. Let’s say that we want the square root of 3.13. That will be our number. Half of that is 1.565. That will be our guess.

IterationResult of EquationResult x Result

In the first iteration, the guess is 1.565. In each successive iteration, the guess used is the result of the last operation. Thus in Iteration 2 the guess is 1.7825. And so on. We stop as soon as result times result is our number.

The bigger the number, the more iterations. The square root of my year of birth (1936) is 44. That took 9 iterations. The square root of 9,876,543.21 is 3142.696804—and that took 15 rounds of what are simply three of the four basic arithmetic operations of multiplication, division, addition, and subtraction.

If you absolutely insist on doing things by hand, it is easy to automate square root pulling in Visual Basic or whatever language you prefer. And such a subroutine could be built into the code I’ve devised (and point to in the earlier posting) to make the operation as “pure” as possible.


  1. There's another process that uses the formula : (a+b)^2=a^2+2ab+b^2

    1. You need to supply what a and b are, Sayantan. Suppose I want the square root of 1936...?

  2. I like the method of pulling roots you have described here. The Japanese (and/or German) methods that you describe in other posts are great as well but I find this one is simpler and it can easily be modified to work to the n'th root:

    The formula can just be re-written as:

    nth root = 1/n * [(number/guess^(n-1) + (n-1)*guess]

    The second column would then be (result of equation)^n

    Fewer iterations are required if the guess is somewhat close to the actual root.

    I haven't discovered how to do n'th roots with the Japanese method.