bounds of spectral radius of hadamard product

176 Views Asked by At

In Topics in matrix analysis by R. A. Horn, C. J. Johnson it's shown that for two positive matrices $A$ and $B$, $\rho(A\circ B)<\rho(A)\rho(B)$. I'm wondering whether the extension of this to positive linear operators is an established result. And is there any restrictions I should worry about?

For instance, $T_1(f)=\int k_1(x,y)f(y)dy$, $T_2(f)=\int k_2(x,y)f(y)dy$ for some strictly positive, square integrable kernels $k_1$ and $k_2$, do we naturally have $\rho(T_1\circ T_2)<\rho(T_1)\rho(T_2)$ where $T_1\circ T_2(f)=\int k_1(x,y)k_2(x,y)f(y)dy$