Why there is no Turing Machine that accepts the Diagonal Language?


Given the diagonal language

$ $ L_d = {i: \sigma_i \notin L(M_i)}$ $

Where $ M_i$ are all Turing Machines and $ \sigma_i$ are all the words, if you put in in a Matrix like this:

$ $ \begin{array} {|c|c|c|c|c|c|c|} \hline & \sigma_1 & \sigma_2 & \sigma_3 & \sigma_4 & \sigma_5 & …\ \hline M_1 & 1 & 0 & 1 & \dotsb & \dotsb & \dotsb \ \hline M_2 & 0 & 0 & 1 & \dotsb & \dotsb & \dotsb \ \hline M_3 & 1 & 0 & 1 & \dotsb & \dotsb & \dotsb \ \hline M_4 & \vdots & \vdots & \vdots & 1 & \dotsb & \dotsb \ \hline M_5 & \vdots & \vdots & \vdots & \vdots & \ddots & \dotsb \ \hline \vdots & \vdots & \vdots & \vdots & \vdots & \vdots & \ddots \ \hline \end{array}$ $

Then $ L_d$ is represented by the numbers in the diagonal of the matrix. In class I was told that there is no TM that accept $ L_d$ , but I do not quite understand why is that, could somebody help?

PS: The above explanation was included because I did not know if this is called Diagonal Language in English, Spanish is my mother tongue.