>Although the term sqrt(ma-Ea) is purely imaginary, for some parameters the solution is purely real:
>
>(%i26) sol:solve([pasquare*pbsquare*costhetavkv=((-ma^2-mb^2+mc^2)/2 + Ea*Eb)^2,pasquare=Ea^2-
>ma^2,pbsquare=Eb^2-mb^2],
> [Eb,pbsquare,pasquare])$
>
>(%i29) subst([Ea=2,ma=1,costhetavkv=0],sol);
>
>(%o29)[[Eb=(-2*mc^2+2*mb^2+2)/8,pbsquare=-(-4*mc^4+(8*mb^2+8)*mc^2-4*mb^4+56*mb^2-4)/64,pasquare=3],
>[Eb=-(2*mc^2-2*mb^2-2)/8,pbsquare=-(-4*mc^4+(8*mb^2+8)*mc^2-4*mb^4+56*mb^2-4)/64,pasquare=3]]
>
>That looks real to me. You might try using the function 'rootscontract.'
Thanks, rootscontract solved problem, but i think that disabling of sqrt(a*b) --> sqrt(a)*sqrt(b) in solve() is better solution.
may be it is possible?