(another comment, not an answer.) Experimentally, with the following code ```python from sympy import * var('x') var('n', integer = True) f = ( (2+(x+3) * sqrt(-x))/(2*(x+4)) * sqrt(-x)**n + ((2-(x+3) * sqrt(-x)) / (2*(x+4))) * (-sqrt(-x))**n + + (x+2+sqrt(x*(x+4)))/(2*(x+4)) * ( (x+sqrt(x*(x+4)))/2 )**n + (x+2-sqrt(x*(x+4)))/(2*(x+4)) * ( (x-sqrt(x*(x+4)))/2 )**n ) pprint(f) for i in range(10): eq = simplify(f.subs(n,i)) print ('========== n = ', i) pprint(eq) print ('have solutions') sols = solve(eq) pprint (sols) pprint ('approx. = ') pprint ([s.evalf() for s in sols]) ``` we see the following: $$ \begin{array}{l} f_{0} = 1\\ f_{1} = 0\\ f_{2} = x^{2}\\ f_{3} = x^{2} \left(x + 2\right)\\ f_{4} = x^{2} \left(x^{2} + 2 x + 1\right)\\ f_{5} = x^{3} \left(x^{2} + 3 x + 1\right)\\ f_{6} = x^{4} \left(x^{2} + 4 x + 4\right)\\ f_{7} = x^{4} \left(x^{3} + 5 x^{2} + 7 x + 3\right)\\ f_{8} = x^{4} \left(x^{4} + 6 x^{3} + 11 x^{2} + 6 x + 1\right)\\ f_{9} = x^{5} \left(x^{4} + 7 x^{3} + 16 x^{2} + 13 x + 2\right)\\ \end{array} $$ [![enter image description here][1]][1] And it seems that already $f_9$ have some imaginary roots, albeit very small. Do these imaginary roots really converges to 0 when $n\to \infty$ or they stay on the same magnitude ? [1]: https://i.sstatic.net/z91X1.png