# Complex roots

1. Jan 31, 2017

### fishturtle1

1. The problem statement, all variables and given/known data
Solve for the roots of the following.
(What do you notice about the complex roots?)

b) x3 + x2 + 2x + 1 = 0

2. Relevant equations
To find roots of a polynomial of degree n > 3, look at the constant and take all its factors. Those are possible roots. Then plug them in to see which ones solve the equation.

cubic equations from this link: http://www.mathportal.org/formulas/algebra/solalgebric.php

3. The attempt at a solution
factors of 1: 1, -1

x=1
(1)3 + 12 + 2(1) + 1 = 0

1 + 1 + 2 + 1 = 0
5 = 0, so 1 is not a root.

x=-1
(-1)3 + (-1)2 + 2(-1) + 1 = 0

-1 + 1 -2 + 1 = 0 -> -1 = 0
So -1 is not a root.

I also tried to use cubic formulas and did not find the correct answer. I checked my calculations.

So i think there is some method more simple than cubic equations that I cannot think of/find on the web.

Are there any other ways besides testing the constant term's factors, and using the cubic equations, to find the roots of a 3rd degree polynomial?

2. Jan 31, 2017

### Ray Vickson

Show us what you got when you used the cubic formulas. Perhaps you made an error, but we cannot tell if you won't show us your work.

3. Jan 31, 2017

### FactChecker

Testing factors of the constant term will almost certainly fail. It's not worth trying unless the problem function has been specially set up for that to work. You should show us how you used the cubic formulas. They should work every time.

4. Feb 1, 2017

### fishturtle1

Sorry, here is my work:

x3 + x2 + 2x + 1 = 0
So, a1 = 1, a2 = 2, a3 = 1

Then Q = (3a2 - a21)/9
Q = (3(2) - (1)2)/9
Q = (6-1)/9 = 5/9

R = (9a1a2 - 27a3 - 2a13) / 54
R = (9(1)(2) - 27(1) - 2(1)3)/54
R = (18 - 27 - 2) / 54
R = -11/54

S = (R + (Q3 + R2)(1/2))(1/3)
S = ( (-11/54) + ( (125/729) + (121/2916) )(1/2) ) (1/3)

T = ( (R - (Q3 - R2)(1/2) ) (1/3)
T = ( (-11/54) - ( (125/729) - (121/2916) )1/2 ) 1/3
T = about 0.82632i, actually I just checked T = about 0.74690

So for my first root i did this:

x1 = S + T + a1/3 = 0.63642 + 0.82632i - 1/3

x1 = 0.30309 + 0.82632i
So with the new T, x1 = 0.30309 + 0.74690i.. but this is still wrong.
at which point I checked with an online calculator and this was not any of the three roots so I'm wondering if I used this formula incorrectly or just made a calculation error.

x1 = -0.56984

x2 = -0.21508 + 1.30713i

x3 = -0.21508 - 1.30713i

5. Feb 1, 2017

### FactChecker

For T, you should use the negative real third root of (R - (Q3 + R2)(1/2)). That is T=-0.872931847948311.
Your values of Q, R, and S are correct.
With the corrected value of T, you should get the correct roots.

6. Feb 1, 2017

### Ray Vickson

For intermediate result you should keep as many digits as your calculator will hold. If you round intermediate quantities to values like 0.74690, etc., you can introduce significant roundoff errors, perhaps even making the final answer useless. (Of course I realize you may have done this, but it is not clear from what you wrote.)

Before embarking on an equation-solving expedition it is useful to first gain some insight into the nature of the problem. First, since all terms are > 0 for positive x, there is no positive root. Descarte's Rule of signs implies that the number of negative roots x < 0 is either three or one, and plotting the function makes it clear that there is only one. Furthermore, the function is > 0 at x = 0 and is < 0 at x =1, so there is a root between -1 and 0.

Almost nobody I know would ever bother using the cubic solution formula, even they know very well what it is; most people would just solve the equation numerically, which you can do to as many digits of accuracy as you could ever want using modern tools.

7. Feb 2, 2017

### fishturtle1

So is that a general rule when using these equations? That you leave the " i " out in the T? and yeah i see that this would help me get the real root. Thank you.

I see what you mean. I remember in my physics class they taught us to never round too much in the in between steps and i forgot to use that here.

I feel like i have forgotten something from pre calculus class but what do you mean by "most people would just solve the equation numerically".. is there a method besides the cubic equations and rational roots test?

8. Feb 2, 2017

### FactChecker

If I was writing a program, I would use the formulas. Numerical solutions can have problems converging and can require multiple attempts (with different initial values) to get a solution.

9. Feb 2, 2017

### FactChecker

Yes. They want T to be a real number. There is always one real cube root of any real number. If you include 'i', that is not a cube root of the number.

10. Feb 2, 2017

### Ray Vickson

The issue is not at all straightforward, due in part to the effects of roundoff errors. (Even programming the quadratic solution robustly is not as easy as one might think.)

There seems to be a debate among the experts as to whether iterative methods (Newton, etc.) or algebraic methods are better; both suffer from problems in implementation. However, there are newer iterative methods available that do not require "good" starting points, but they are more sophisticated than straight Newton-Raphson.

The OP might like to consult
http://math.stackexchange.com/quest...ng-cubic-and-quartic-equations-by-computer-pr
or
http://stackoverflow.com/questions/2003465/fastest-numerical-solution-of-a-real-cubic-polynomial

Some contributors to the latter recommend solving a cubic by setting up the companion matrix (for which the cubic is the characteristic polynomial) then using modern numerical linear algebra methods to find the eigenvalues of the matrix; they claim that that is how some of the available solvers actually work.

All of this matters only if you want to write your own code for solving the problem; otherwise (under what I call modern methods), just use one of the many, many cubic (or higher) solvers available already. For example, Wolfram Alpha works well and is free.

11. Feb 7, 2017

### fishturtle1

Thanks for all your thoughtful answers. This is a late response because after reading some of those articles I was felt out of my depth and decided to come back to this problem set after practicing some easier problems.

Funny thing is, the instructor said the problem was a typo...it was supposed to read x3 + x2 + x + 1 = 0..

which I can see x = -1 is a root

so I can factor out x-1 and rewrite the original equation as: (x+1)(x2+1) = 0

then solve x2 + 1 = 0

x2 = -1

x = i, x = -i

So the roots are x = -1, -i, and i...

now i will mark this as solved..just wanted to add the solution incase someone else wanted it.