1/gamma(k) can be handled per the link in my previous post.
But u*u' is non-convex quadratic, and doesn’t get better when divided by gamma.
Have you proven this is convex? Maybe the intention is to have a convex relaxation U \succeq u*u', and use U in place of u*u' in Q? But then what happens when dividing by gamma(k)? Or maybe there’s (also) some kind of SCA thing. But it’s your problem, in the sense that perhaps you lifted it from a paper or book, so you need to understand it, and figure out what the convex optimization problem, if any, is.