Reformulation of Goldbach's Conjecture as optimization problem correct?

255 Views Asked by At

Question

I think I managed to reformulate a stronger version of Goldbach's conjecture as an optimization problem:

$$ \frac{\partial F_n}{\partial a_n} = \frac{\partial F_n}{\partial \overline a_n} = \frac{\partial F_n }{\partial \lambda_n}=0$$

Where:

$$ F_n = |a_n|^2 - \lambda_n \frac{\sin(\pi |b_n|^2)}{|b_n|^2} $$

$$ (\sum_{r=1}^\infty |a_r|^2 x^{2r+1})^2 = \sum_{r=1}^\infty |b_r|^2 x^{2r+4}$$

Is this formulation correct? Is this stronger version convincing? Does it have any interesting implications?

Background

We are all familiar with Goldbach's Conjecture any even number greater then $2$ is expressible as the sum of primes. Expressing it combinatorically (without the case of $p_1$):

$$ (\sum_{r=2}^\infty x^{p_r})^2 = \sum_{r=1}^\infty b_r x^{2r+4} $$

where, $p_r$ is the r'th prime and $ b_r \neq 0 $.

I recently thought that a stronger version as a limit of $n \to \infty$:

"The minimum number of elements in a set required to construct the even numbers $<n$ as a sum of it's elements is the primes numbers"

Expressing the above as equations:

$$ \min(|a_r|^2) = \begin{cases} 1 \hfill & 2r+1=\text{prime number} \\ 0 \hfill & \text{ otherwise} \\ \end{cases} $$

Under the constraints:

$$ (\sum_{r=1}^\infty |a_r|^2 x^{2r+1})^2 = \sum_{r=1}^\infty |b_r|^2 x^{2r+4}$$

and, $b_r$ must be an a natural number

$$ \frac{\sin(\pi |b_n|^2)}{|b_n|^2} = 0 $$

For example

By comparing coefficients of $x$:

$|a_1|^4=|b_1|^2$

Using Lagrange multipliers:

$$ F_1= |a_1|^2 - \lambda_1 \frac{\sin(\pi |a_1|^4)}{|a_1|^4} = a_1 \overline a_1 - \lambda_1 \frac{\sin(\pi (a_1 \overline a_1)^2)}{(a_1 \overline a_1)^2} $$

Where $ \overline a_n $ is the complex conjugate of $a_n$.

Now, finding the minimum

$$ \frac{\partial F_1}{\partial a_1} = \frac{\partial F_1}{\partial \overline a_1} = \frac{\partial F_1 }{\partial \lambda_1}=0$$

This gives us the following equations: $$ \overline a_1 - \lambda_1 ( 2 \pi \frac{\cos(\pi( a_1 \overline a_1)^2)}{ a_1} - 2 \frac{\sin(\pi (a_1 \overline a_1)^2)}{(a_1^2) (\overline a_1)^3} ) = 0 $$

$$ a_1 - \lambda_1 ( 2 \pi \frac{\cos(\pi( a_1 \overline a_1)^2)}{ \overline a_1} - 2 \frac{\sin(\pi (a_1 \overline a_1)^2)}{(\overline a_1^2) ( a_1)^3} ) = 0 $$ $$ \frac{\sin(\pi (a_1 \overline a_1)^2)}{(a_1 \overline a_1)^2} = 0 $$

Solving the above:

$ |a_1|^2=1$ and $ \lambda_1 = -\frac{1}{2 \pi} $