Error displayed in
tU * (a0 * trace(GU_hat * F2) + 1) + log(tU) + 1 …
^
tU * (a0 * trace(GU_hat * F2) + 1)
is a linear -fractional form, as is the next line.
Given the stuff in the rest of the problem, I don’t know whether this can be transformed into a problem CVX would accept using the technique, or similar, of section 4.3.2 of https://stanford.edu/~boyd/cvxbook/bv_cvxbook.pdf . I don’t even know whether this is a convex optimization problem (or convertible thereto). But you should.