> Achieving a truly collimated beam where the divergence is 0 is not possible, but achieving an approximately collimated beam by either minimizing the divergence or maximizing the distance between the point of observation and the nearest beam waist is possible.
The diffraction limit is a well supported theoretical limit that is caused by the wave nature of light. There is ways to get around it in the near field (up to a few ten wavelength away from the source). But in the far field you can not get a perfectly divergence-free finite-width beam. [1] Nor can you focus a beam to a beam waist smaller than about the wavelength.
In practice the situation is even worse and you can't even get the performance that a perfect Gaussian beam would allow. We often express the performance of real beams by a thing called M^2 or beam quality parameter [2]. In some sense is measures how much wider the beam is than it needs to be, and this number is never less than one.
> Achieving a truly collimated beam where the divergence is 0 is not possible, but achieving an approximately collimated beam by either minimizing the divergence or maximizing the distance between the point of observation and the nearest beam waist is possible.